\documentclass{article}
\usepackage[utf8]{inputenc}
\usepackage[english]{babel}
\usepackage[T1]{fontenc}
\usepackage[affil-it]{authblk}
\usepackage{csquotes}
\usepackage{appendix}
\usepackage[colorlinks,linkcolor=black,citecolor=cyan,urlcolor=blue,bookmarks=false,hypertexnames=true]{hyperref} 
\usepackage[style=authoryear,bibstyle=authoryear,citestyle=apa,natbib=true,hyperref=true,abbreviate=true]{biblatex}
\usepackage{setspace}
\usepackage{multirow}
\usepackage{array}
\usepackage{amssymb}
\usepackage{dirtytalk}
\usepackage{graphicx}
\usepackage{rotating}
\usepackage{caption}
\usepackage{subcaption}
\usepackage{enumitem}
\usepackage{amssymb}
\usepackage{calc}
\usepackage{adjustbox}
\usepackage{float}
\usepackage{filecontents}
\usepackage{geometry}
\usepackage{titling}
\setlength{\droptitle}{-0.5in}
\usepackage{tablefootnote}
\usepackage{chngcntr}
\usepackage{pdflscape}
\setlength{\parskip}{1em}
\renewcommand{\baselinestretch}{1.6}
\newcommand{\subscript}[2]{$#1 _ #2$}
\newcommand{\ndd}{Unpublished}
\newcommand{\xmark}{\ding{55}}
\DeclareNameFormat{labelname:poss}{% Based on labelname from biblatex.def
  \nameparts{#1}% Not needed if using Biblatex 3.4
  \ifcase\value{uniquename}%
    \usebibmacro{name:family}{\namepartfamily}{\namepartgiven}{\namepartprefix}{\namepartsuffix}%
  \or
    \ifuseprefix
      {\usebibmacro{name:first-last}{\namepartfamily}{\namepartgiveni}{\namepartprefix}{\namepartsuffixi}}
      {\usebibmacro{name:first-last}{\namepartfamily}{\namepartgiveni}{\namepartprefixi}{\namepartsuffixi}}%
  \or
    \usebibmacro{name:first-last}{\namepartfamily}{\namepartgiven}{\namepartprefix}{\namepartsuffix}%
  \fi
  \usebibmacro{name:andothers}%
  \ifnumequal{\value{listcount}}{\value{liststop}}{'s}{}}
\DeclareFieldFormat{shorthand:poss}{%
  \ifnameundef{labelname}{#1's}{#1}}
\DeclareFieldFormat{citetitle:poss}{\mkbibemph{#1}'s}
\DeclareFieldFormat{label:poss}{#1's}
\newrobustcmd*{\posscitealias}{%
  \AtNextCite{%
    \DeclareNameAlias{labelname}{labelname:poss}%
    \DeclareFieldAlias{shorthand}{shorthand:poss}%
    \DeclareFieldAlias{citetitle}{citetitle:poss}%
    \DeclareFieldAlias{label}{label:poss}}}
\newrobustcmd*{\posscite}{%
  \posscitealias%
  \textcite}
\newrobustcmd*{\Posscite}{\bibsentence\posscite}
\newrobustcmd*{\posscites}{%
  \posscitealias%
  \textcites}
\addbibresource{references.bib}
\title{Territorial Loss and Nationalist Populism \\
\let\thefootnote\relax\footnotetext{\onehalfspacing\vspace{-.35cm}\\ %Carter: Postdoctoral Associate, New York University. \href{mailto:pjc504@nyu.edu}{pjc504@nyu.edu}.  \\ Pop-Eleches:  Professor of Politics and International Affairs, Princeton University. \href{mailto:gpop@princeton.edu}{gpop@princeton.edu}. \\ 
Replication materials and code can be found at \citet{DVN/QBI7AD_2024}\-\hspace{.5cm}
}}
\author{\vspace{-5em}}
%\author{Perry Jess Carter \and Grigore Pop-Eleches}
\date{\today}

\begin{document}

\maketitle
\begin{abstract}
This paper highlights the overlooked role of prior grievances stemming from historical territorial loss as a significant factor behind support for nationalist populist parties. While not essential for the emergence of nationalist populism, territorial loss uniquely aligns with the backward-looking victimization framing crucial to these parties' electoral success. Utilizing cross-national experimental and observational data from original surveys conducted in 2020–2021 in Romania, Hungary, Germany, and Turkey, we establish territorial loss attitudes as a robust predictor of nationalist populist party support. In addition to variations in national context, a trade-off emerges for governing populists, revealing that priming past losses attracts concerned voters but alienates those indifferent to territorial issues. Analyzing a quasi-natural experiment involving a new nationalist populist party that emerged in Romania between waves, we conclude that loss attitudes are stable over time and temporally prior to support for nationalist populism.

    % In this paper, we argue for the importance of a previously-overlooked factor -- the existence of prior grievances over the historical loss of territory -- as a source of support for nationalist populist parties. While territorial loss is not necessary for the emergence of populism, the issue lends itself unusually well to the backward-looking, loss-oriented framing of national victimization at the hands of elites that constitutes a key element of such parties' electoral success. Drawing on original cross-national experimental and observational data from surveys conducted in Romania, Hungary, Germany, and Turkey in 2020--2021, we demonstrate that territorial loss attitudes are a remarkably robust predictor of support for nationalist populist parties. However, important differences in national context emerge across cases. Moreover, we show the existence of a trade-off that only faces populists in power, where priming past losses alienates voters who are unconcerned about territorial loss even as it attracts those for whom it is important. In addition, the panel structure of our data allows us to exploit a quasi-natural experiment in the form of the emergence of a new and highly successful nationalist populist party in Romania between waves, from which we conclude that loss attitudes are stable over time and temporally prior to support for populism.
\end{abstract}
\section{Introduction}
\begin{flushleft}

The rise of nationalist populist parties in much of Europe in the past decade \citep{vachudova2020ethnopopulism,vachudova2021populism,hunger2022s} has attracted considerable scholarly attention, most recently in the context of resistance to public health measures during the COVID-19 pandemic \citep{bobba2021populism,eberl2021populism,kaltwasser2022populist}. Existing explanations have focused on the role of immigration \citep{hobolt2016fleeing,halikiopoulou2020economic}, corruption \citep{engler2020fighting,snegovaya2020voice}, traditional values \citep{taggart2000populism}, the threat of Islam \citep{brubaker2017between}, or marginalization \citep{gidron2020populism,noury2020identity,adler2021struggles}, others on economic policy choices, such as the embrace of neoliberal economic policies by leftist parties \citep{berman2019populism} or the consequences of neoliberal reforms more broadly \citep{appel2018triumph}.

This paper focuses on a different -- and so far largely ignored -- source of support for nationalist populist parties: the politics of territorial losses . While the role of irredentist claims in driving conflict has received significant attention in the international relations literature \citep{chazan1991irredentism,saideman2000determining,toft2014territory,shelef2016unequal}, the issue has received much less attention in the debates about the drivers of nationalist populist party success. Partial exceptions in this regard are \citet{pirro2014populist}, who discusses irredentist rhetoric as part of the broader repertoire of East European radical right parties and \citet{heinisch2019people}, who analyze the ``territorial'' dimension of populism in the context of latent regional center-periphery cleavages. However, to the best of our knowledge the impact of irredentist territorial claims on the support for nationalist populist parties has not been systematically theorized and tested to date.

Nevertheless, there are strong empirical and theoretical reasons to expect that, where present, such claims frequently form an integral and effective part of populist -- specifically, nationalist populist -- political appeals. Explicit calls for military action to regain lost territory in democratic systems are, of course, quite rare. More general discussions of lost territory, ranging from post-imperial nostalgia to lamentations of victimization by external foes and an indolent political class, are, however, frequently heard from nationalist populists, even in cases such as contemporary Germany where irredentism \textit{per se} is widely discredited \citep{jackson2019germany}. Such messaging is especially common in the context of commemoration of important historical events. Highly typical in this regard is the following passage from a 2018 statement by Hungarian Prime Minister Viktor Orb\'{a}n commemorating the failed 1848 Hungarian Revolution, a major national holiday:

\begin{quote}
    \textit{The situation, my dear friends, is that they want to take our country. Not with a stroke of a pen, like a hundred years ago in Trianon. Now they want us to voluntarily hand it over in a few decades to others, to strangers from other continents who do not speak our language, do not respect our culture, laws and way of life, who want to replace ours with their own (\cite{orban2018}).}
\end{quote}

Orb\'{a}n's rhetorical framing makes explicit the association between past territorial grievances and the bread-and-butter issues of present day nationalist populists. Drawing a direct analogy between the foreign `other' responsible for physically taking parts of the country away in the past and their supposed modern incarnations who now promote open borders, the issues of today are elevated to the status of a century-long campaign of victimization. Commemorating the 80th anniversary of the Soviet occupation of parts of Romania, the tellingly-named Alliance for the Union of Romanians (AUR), an ultra-nationalist populist party that rose to prominence in Romania during the COVID-19 pandemic, made a still more emotionally resonant appeal, weaving together messianic narratives with critiques of both domestic and foreign elites:


\begin{quote}

\textit{One Friday, Christ was unjustly crucified, He, the one without fault.}

\textit{One Friday, Bessarabia, Northern Bukovina and Herța County were ceded, approximately 18\% of the surface of Romania at that time, territories inhabited by 4 million people, mostly Romanian ethnics.}

\textit{In the face of a brutal and unjust ultimatum from the USSR, a country whose leaders claimed to be the most progressive form of social order, a Romanian political class incapable of greatness and honor gave its consent to an ignominy.}

\ldots

\textit{The politicians who control the Romanian state today will not organize any demonstration, will not give any speech and will not commemorate in any way this unfortunate moment of our history. People cut from the same fabric of cowardice and as petty in soul as those who ruled Romania 80 years ago will ignore June 28th. Their thoughts are on the acquisitions to be made, on the commissions that can be extracted, on the upcoming elections in which the people must again be deceived.}

\ldots

\textit{\textbf{AUR will act to correct the consequences of the territorial and human abduction produced on June 28, 1940}, will support economically and culturally those Romanians who remained outside the country's borders so that they can preserve their national identity, and will take the necessary steps at a political and diplomatic level to achieve the union of the two Romanian states.} (\cite{party_2020}, emphasis original)
\end{quote}

As we argue in this paper, this apparent affinity between nationalist populism and the politicization of territorial loss is not coincidental, but results from a combination of the issue's particular symbolic suitability for the type of backward-looking and territorially grounded narratives such parties typically advance and its tendency to be ignored by mainstream political actors. In general, the drawing of territorial boundaries is fundamental to the enterprise of the historical construction of the nation.  In \posscite[p. 1]{gellner2008nations} classic formulation, the principle of nationalism \textit{by definition} \textquote{requires that ethnic boundaries should not cut across political ones} .  Gellner's emphasis on \textit{ethnic} boundaries is, of course, problematic in the many cases in which membership in the nation is not defined in purely ethnic terms, or where the imagined geographic reach of the nation includes land not currently occupied by members of the national community.  Nevertheless, the fact remains that the definition of the physical boundaries of the nation, and highlighting instances of their incongruence with the state, is precisely the stuff that nationalist politics is made of. \par
For this reason, political entrepreneurs seeking to mobilize support on a nationalist basis enjoy greater symbolic resources in those countries where the mismatch between the actual and imagined boundaries of the nation is the greatest; that is, \textit{where control over a substantial amount of territory widely believed to be part of the nation has been lost to another state}.  Indeed, a glance at Eastern Europe is consistent with this intuition: Poland and Hungary, which lost large proportions of their territory during the 20th century, have experienced some of the highest support for nationalist populist parties in the region despite their otherwise successful post-communist trajectories. \par

The relationship between territorial loss and nationalist populism, we contend, goes beyond the general importance of boundary drawing to nationalist movements. Rather, the issues around which most nationalist populist narratives center, such as opposition to immigration, trade liberalization, or supranational integration projects, share a common core of features in addition to their obvious nationalist content \citep{bonikowski2019populism,halikiopoulou2019new,halikiopoulou2020economic}. To a greater or lesser degree, almost all of these issues are framed in terms that \textit{look backward} to a (usually imagined) past that compares favorably with the present and emphasize \textit{loss} and \textit{victimization} at the hands of external actors -- typically foreign and domestic \say{elites} \citep{gest2018roots,elcci2022politics}. The issue of lost territory, while partially dependent on the specific circumstances of the loss, powerfully combines these features, lending itself to the mobilization of emotionally resonant narratives of national victimization and nostalgia for better times consistent with the broader nationalist populist worldview.  Indeed, a core feature of populist narratives more generally is the appeal to a deliberately vaguely-defined national ``heartland'' \citep{taggart2004populism} in which the ``true'' people reside. This resonates closely with irredentist appeals to homeland \citep{shelef2020homelands} or past golden ages \citep{cederman2024future}, which allow political entrepreneurs to speak on behalf of a people and a place that may not even exist in contemporary reality.

At the same time, the politicization of this issue is in most cases unavailable to mainstream political elites and especially to current incumbents. Due to the futility of most attempts to recover lost territory and the widespread international norms of respecting existing national borders, sustained mobilization of claims on lost territory in competitive political environments, even if purely symbolic, invites discontent over a lack of results \citep{shelef2020homelands}. This is especially true in contemporary Europe, where international integration has severely limited opportunities for actively pursuing territorial claims. As a result, existing popular concern over historical losses -- which in most cases far predates the emergence of \say{new} issues such as mass immigration -- frequently goes unattended by the political mainstream, making it a ripe issue for mobilization by non-systemic parties. Nationalist populist parties in power, however, may face the same constraints as mainstream parties, underscoring the need to distinguish between populism in opposition and in government \citep{albertazzi2015populists,pappas2019populists,zellman2019hawking,aytacc2021partisanship,kaltwasser2022populist,guriev2022political,funke2023populist}. At worst, they may provoke a double backlash: dissatisfaction with the lack of progress from the most ardent irredentists, and irritation from those who prefer to leave historical losses in the past.

We test this argument using data from an original survey conducted in July 2020 in Romania, Hungary, Turkey and Germany, including an embedded question order experiment that allows us to estimate the effect of priming the salience of territorial loss on party evaluations while retaining unbiased measurements of loss attitudes for all respondents. Our findings strongly support a robust association between territorial loss attitudes and support for nationalist populist parties, particularly those in opposition, that is comparable in magnitude to other well-established predictors of populist support. At the same time, we find only weak evidence that increases in the salience of territorial loss led to more positive evaluations of populists. This overall non-effect is partly driven by heterogeneous effects of the prime on voters who place more or less importance on lost territory, with the positive effect among high-concern voters offset by a backlash from low-concern voters. In addition, we provide behavioral evidence strongly indicating a causal effect of territorial attitudes on actual voting outcomes using a follow-up panel survey conducted in Romania after the December 2020 parliamentary elections, in which the nationalist populist AUR party participated for the first time.  

These findings contribute to what \citet{kaltwasser2017oxford} and \citet{mudde2018studying} refer to as the \textit{ideational approach} to the study of populism, highlighting the potential for historical grievances based on territory to form an additional foundation for the ``thin'' anti-elite ideology espoused by nationalist populists. Our argument is particularly relevant to the recent literature on nostalgia as a source of populist support \citep{gest2018roots,stefaniak2021different}, as we demonstrate that populists benefit not only from those who perceive themselves or those like them to have lost status \textit{within society} over time, but also from those who are concerned about physical losses incurred by the nation as a whole in comparison to an imagined past.


\section{Populism and Territorial Loss}
\subsection{Sources of Populism}

Although populism is a notoriously poorly defined concept \citep{van2014populist}, the dramatic rise in the electoral success of parties that exhibit populist characteristics, particularly in Europe, over the last decade has attracted a great deal of academic attention \citep{moffitt2016global,algan2017european}. Drawing on definitions by \citet{mudde2017populism} and \citet{norris2020measuring}, by the term \say{populism} we mean primarily an \textit{ideational} feature of the \textit{language} employed by political actors, distinguished by the setting up of rhetorical opposition between the \say{people} and \say{elites} and the idea that the will of the nation ought to be decisive in decision-making. 

While populism, in this sense, needs in principle neither be a stable feature of specific parties nor correlate with the left-right dimension \citep{pop2010throwing}, it has recently been most prominent among the extreme right, especially in advanced industrial democracies \citep{pelinka2013right,norris2019cultural,norris2020measuring}. For this reason, while many of the parties studied in this article are generally considered to be right-wing, we focus specifically on instances of \textit{nationalist populism}. Although \citet{de2017populism,de2017distinctions,de2020should} have argued in favor of disentangling nationalism and populism in order to analyze them as independent concepts, we concur with \citet{brubaker2020populism} in his response that in practice they are too closely intertwined to fully separate. Although populism and nationalism share a basic core idea of elevating the ``people'' and drawing boundaries between \textit{us} and \textit{them}, populism is defined primarily by its discursive and mobilizational style and opposition to elites \citep{mudde2017populism,norris2020measuring,brubaker2020populism} while nationalism can easily be both mundane and elite-led \citep{varshney2021populism}. Although such ``conventional'' nationalists may have little to gain from adopting populist rhetorical styles, however, populists can benefit greatly from ``fusing ... the moral frame of populism with the affective power of nationalism'' \citep[~p.260]{singh2021populism} by superimposing the populist ingroup-outgroup structure of \textit{people} vs. \textit{elites} onto the existing template of the \textit{nation} and its \textit{enemies}. 

As \citet{singh2021populism} argues, this frequent collocation of nationalism and populism tends to enhance the existing tendencies of populist rhetoric, charging the habitual animosity towards ``them'' with the appellation of national traitors while amplifying the benefits accorded to ``us''. Importantly for our application, it also transfers the usual arena of populist rhetoric -- the domestic struggle between people and elites -- into one that transcends state borders, involving supranational organizations as well as old enemies of the nation. 

In this article, we therefore focus on the relationship between this combination of nationalism and populism and lost territory, rather than attempting to separate out the effects of ``pure'' populism. This conceptual entanglement is also reflected in an empirical tendency for populists to become \textit{nationalist} populists. Of the 9 populist parties across 4 countries considered here, only one (\textit{Die Linke} in Germany) does not also qualify as nationalist.

An alternative, but closely related, conception is what \citet{vachudova2020ethnopopulism} and \citet{jenne2021populism} have characterized as \textit{ethnopopulism}. In addition to the anti-elite orientation characteristic of populism more generally, ethnopopulism is characterized by a flexible approach to the definition of an ethnic nation or \textit{people} that requires defense from an array of -- often fabricated -- internal and external enemies \citep{bonikowski2017ethno,vachudova2020ethnopopulism,vachudova2021populism,jenne2021populism}. Such parties are typically characterized by a traditionalist orientation and an opposition to globalization and modernization processes, particularly with respect to immigration \citep{pelinka2013right}. Although this category is closely related to nationalist populism, we choose instead to focus on \textit{nationalist} populism more broadly conceived, since the phenomenon under consideration also applies to nations with a nonethnic basis. For instance, populists may invoke imperial nostalgia to capitalize on a desire to return lost imperial territories in spite of the absence of any common ethnic basis.

Various explanations for the rise in support for these parties have been offered, most of which have focused on economic, cultural, or political factors. On the economic side, scholars have emphasized that support for populist parties comes mainly from the economic \say{losers} of globalization, liberalization, and modernization processes, which tend to seek alternative solutions to their relative losses \citep{pelinka2013right,appel2018triumph,berman2019populism,rodrik2021does}. International economic processes are particularly decisive, with localized exposure to trade shocks strongly driving support for far-right populist parties \citep{milner2021voting}. 

A second set of explanations focuses on value change and cultural threat as driving factors behind individual support for populism. \citet{norris2019cultural}, for instance, find consistent support for \say{cultural backlash} against the increasing adoption of progressive values by the political mainstream as a driving factor behind the defection to populist alternatives, who promise to defend \say{traditional} values \citep{taggart2000populism,brubaker2017between,noury2020identity}. Hostility toward social \say{others} forms a particularly important component of the cultural dimension of populism, as mere exposure to refugees \citep{dinas2019waking} is sufficient to push voters towards far right parties. At the aggregate level, \citet{bustikova2014} shows that increases in the electoral performance of parties representing domestic minority groups trigger subsequent increases in the vote shares of populist parties, likely driven by status threat \citep{mutz2018status}. 

Finally, support for populist parties may be driven less by attitudinal alignment with specific policies they promote than by general dissatisfaction with the political establishment. Perceptions of corruption among elites are especially important in this regard, as populists are able to capitalize on a preexisting sense that the political establishment is hostile to the people \citep{grzymala2017global,engler2020fighting,snegovaya2020voice}. Dissatisfaction may also be simply performance-based, as repeated failures to provide economic benefits despite alternation of power drive protest voters toward populist alternatives \citep{pop2010throwing}. Similarly, when policies viewed as illegitimate or undesirable are uniformly embraced by mainstream parties, voters are likely to be driven toward anti-systemic parties even without strong value congruence \citep{berman2019populism}.

While the literature just discussed provides considerable insight into the general antecedents of support for populism, a disconnect remains between these findings and the mobilization strategies that populist parties employ. A growing body of findings suggests that economic and cultural factors interact with each other, with background economic resentment increasing susceptibility to identity-based appeals \citep{noury2020identity}. At the same time, studies that focus on the discourse employed by populists have repeatedly noted the key role of appeals relying on historical memory, especially with regard to nostalgia for a better -- and frequently both higher status and more territorially extensive -- past \citep{pirro2014populist,koposov2017memory,gest2018roots,caramani2019national,esteve2022reconquest,elcci2022politics}. Unlike many of the \say{new} cleavages, these issues far predate the current wave of populism, leaving open the question of how they interact with these sources of broader dissatisfaction with mainstream politics and how effective populist appeals based on historically constructed loss are at generating support.  

\subsection{Territory, Loss, and National Identity}
    Although the concept of territory is often treated as relatively unproblematic -- from a measurement standpoint, it is generally straightforward to identify the geographical delimitations of formal political units -- there are many subtle, yet important, barriers to conceptualizing \say{control over territory} in a manner that permits valid comparison. In the sense of an area of land with arbitrarily drawn boundaries, a single territory may be interpreted in many ways by different groups and actors, resulting in multiple subjective constructions of the same physical space. The issue of \textit{which} territory is appropriate to consider is also a complex one, as the boundaries drawn by different visions of \say{homeland} or even \say{region} generally do not overlap \citep{knight1982identity}.  The breakaway republic of South Ossetia, for example, overlaps with multiple administrative divisions of the Republic of Georgia, each with its associated historical meaning, while for many Ossetians, it is part of a larger region of Ossetia-Alania that includes parts of the Russian North Caucasus \citep{SHNIRELMAN2001}.  Moreover, the state-centric frame of ``territorial control'' is often inadequate: states may not have \textit{de facto} control over land that is formally part of their sovereign territory due to the presence of local nonstate actors, and imperial powers often exercise significant control outside of their official boundaries \citep{agnew1994territorial,agnew2017globalization}. \par 
   We resolve these ambiguities by adopting a purely subjective definition of territory that does not privilege the state and allows both boundaries and the fact of control to vary at the individual level. Thus, for instance, the territory of Chechnya in Russia does not necessarily need to coincide with the administrative boundaries of the Chechen Republic or be considered under Russian control.  The answer to these questions depends on the \say{mental geography} \citep{knight1982identity} of each person: some emphasize the formal subordination of the Republic to the Russian Federation and are content to conflate its legal with its actual borders, while others view its extensive autonomy as a sign of non-control or consider parts of neighboring Ingushetia to be Chechen land \citep{cornell2005small}. \par
    While these views result from ongoing processes of social construction, they nonetheless exist at the level of individual perceptions. Without knowledge of the distribution of these perceptions, any attempt to aggregate them into a \say{generally accepted} view would be potentially misleading.  We, therefore, adopt the more conservative approach of allowing each person to define their national geography, although the lack of data on these perceptions will necessitate the imposition of simplifying assumptions when operationalizing this definition (cf. \cite{shelef2016unequal}). This implies that the same objectively defined territory may be simultaneously lost and not lost --  regardless of its empirical status -- depending on individual beliefs about where it is appropriate to draw boundaries and what constitutes control. In extreme cases, for example, some people might consider their entire country to be \say{lost} to an illegitimate or foreign-backed regime despite having legal sovereignty. \par
    Much of the classic work on nationalism emphasizes the centrality of the territorial construction of memory in the development of modern nationalism \citep{gellner2008nations,Smith1986,smith1996culture}. In order to imagine the national community, it is necessary also to associate with a particular place and to draw spatial lines that separate co-nationals from others \citep{anderson2006imagined}.  These authors, however, generally advance a view of nationalism as an ideology that is concerned with the conceptual coherence of nationalist thought but does not convincingly deal with why or how people respond to nationalist appeals. Given the general incoherence of the political ideologies held by most individuals \citep{converse2006nature}, this is a non-trivial issue and points to the need to establish the microlevel mechanisms linking an actual violation of nationalist principles with the emergence of nationalist politics. \par
    In addition, much early work on nationalism defines the nation in ethnocultural terms \citep{gellner2008nations,Smith1986}, leaving ambiguous the processes associated with \say{civic} nations. Territory is as important to nations defined in political terms as cultural terms, to the extent that they exist as a pure type \citep{smith1991national,yack1996myth}. Any theory of the link between territory and nationalism must, therefore, avoid the trap of conflating the nation with ethnicity and focus on general mechanisms associated with the nation as a \textit{group} and not specifically as an \textit{ethnic} group. 
    \par
    More recently, a large literature in international relations has recognized the significance of territory, and its association with nationalism, as a source of conflict. Although much of this literature emphasizes the material value of territory, with inter-state conflict driven by the desire to control strategic and economic resources \citep{gibler2007bordering,huth2009standing}, an important strand emphasizes the peculiar meaning attached to land by nationalists and the potential for bargaining to break down due to claims of indivisibility \citep{toft2010geography,toft2014territory,goddard2006uncommon,shelef2016unequal,manekin_grossman_mitts_2019}. In particular, \citet{shelef2016unequal} finds that the loss of territory that is discursively constructed as a homeland leads to increased conflict at all levels of intensity compared to territory that is not viewed as integral to the nation. However, this literature provides limited insight into the domestic political processes connecting territory to nationalist politics, as these studies tend to reify groups and either treat primordial attachments as given or as reducible to the economic interests of elites. \par 
    In an important recent study, for instance, \citet{manekin_grossman_mitts_2019} employ conjoint experiments to study attachment to Palestinian territory among Israelis at the individual level, but their focus is on understanding trade-offs between control over territory and material values and their effect on elite bargaining positions.  The authors attribute the value placed on territory to partisanship, with supporters of left-wing parties willing to give up land in exchange for prosperity and security, while right-wing voters are strictly opposed to compromise, but the \textit{causal} role of partisan loyalty remains unclear, as do the mechanisms underlying attachment to territory. Similarly, \citet{zellman2019hawking} finds that in Israel and Serbia, only voters with irredentist outlooks responded positively to the symbolic framing of territorial foreign policy goals, concluding that nationalist populist parties are limited in their ability to garner support from the wider electorate. The extent to which this tendency reflects a causal relationship between the prevalence of underlying irredentist attitudes and the success of such parties, however, remains an open question. \par
    The literature discussed thus far treats territory passively -- either as a good to be competed over, or as a symbolic reference point for the construction of a narrative. Land is not simply an empty container, however, instead being closely associated with those who live on it. In particular, when territory is lost, it typically creates a diaspora population either of co-nationals \say{left behind,}, internally displaced persons distributed throughout the losing country, or refugees in third countries. The existence of a significant diaspora living in lost territory significantly complicates the politics of nationalism, as the diaspora constitutes a separate actor with distinct interests and a potential symbolic resource for nationalizing elites \citep{brubaker1995national,brubaker2011nationalizing}. Foreign diasporas, moreover, often intervene actively in the construction of both identity and territoriality, tending to reproduce highly essentialist conceptions of nation and homeland and influence the politics of both the host and home country in support of their agenda \citep{schulz2003palestinian,carter2005geopolitics,homcconnell2019diaspora}.  Consequently, wherever such diaspora populations are present, the relationship of individuals to lost territory cannot be fully accounted for without also considering their relation to the current and former occupants of that territory. \par

\section{Lost Territory as a Nationalist Populist Resource}
\label{sec:theory}

In this section, we present a novel account of the role played by territorial loss as an element of the broader repertoire of nationalist populist rhetoric, explaining why and under what circumstances it is effective at generating support. We begin by considering the common features of the issues typically emphasized by populists and the reasons for their persuasive power, highlighting the particularly fertile ground created by territorial losses in each regard. We then move on to discussing the limitations faced by parties that aspire to govern in effectively mobilizing the issue.

Although, as discussed previously, the specific issues that form the basis of populist party platforms vary considerably across national contexts, they share a set of basic commonalities that closely relate to their persuasive effectiveness: \textit{nationalism}, \textit{loss framing}, \textit{victimization narratives} and \textit{looking backward}. 
% Table \ref{tab:theorymatrix} summarizes the role typically played by each of these dimensions in common populist narratives graphically\footnote{The list of policies is not intended to be comprehensive, nor is the framing of each issue necessarily consistent across all parties: for instance, economic decline may be interpreted as a result of victimization by foreign enemies, as in case of economic sanctions. Instead, Table \ref{tab:theorymatrix} simply reflects the most common framing of some of the most prominent populist issues across cases.}.
We now discuss each in turn.

% \begin{table}[!htbp] \centering
% \caption{A typology of populist issues}
% \label{tab:theorymatrix}
% \resizebox{\textwidth}{!}{
%     \setlength{\extrarowheight}{2pt}
%     \begin{tabular}{cc|c|c|c|c|}
%       & \multicolumn{1}{c}{} & \multicolumn{4}{c}{Feature}\\
%       & \multicolumn{1}{c}{} & \multicolumn{1}{c}{\textit{Nationalist}}  & \multicolumn{1}{c}{\textit{Loss-oriented}}& \multicolumn{1}{c}{\textit{Victimization}}&\multicolumn{1}{c}{\textit{Backward Looking}} \\\cline{3-6}
%       \multirow{5}*{Issue}  & \multirow{1}*{\textit{Economic Decline}} & -- & \checkmark & -- & \checkmark \\\cline{3-6}
%       & \multirow{1}*{\textit{Immigration}} & \checkmark & \checkmark & -- & \checkmark \\\cline{3-6} 
%        & \multirow{1}*{\textit{Social Policy}} & -- & \checkmark & \checkmark & \checkmark \\\cline{3-6} 
%       & \multirow{1}*{\textit{Anti-Globalization}} & \checkmark & \checkmark & \checkmark & \checkmark \\\cline{3-6} 
%       & \multirow{1}*{\textit{Territorial Loss}} & \checkmark & \checkmark & \checkmark & \checkmark \\\cline{3-6} 
%     \end{tabular}
% }
% \vspace{\fill}
% \end{table}

The core feature of nationalist populist narratives is, of course, \textit{nationalism} \citep{de2017populism,vachudova2020ethnopopulism,vachudova2021populism,singh2021populism}. Clearly, not all rhetorically salient issues for such parties are directly tied to nationalist logic. In particular, neither economic grievances nor cultural backlash to progressive social policy have any inherent relationship to nationalist projects. However, even these issues are frequently given a nationalist coloring by framing them as the result of foreign influence undermining the \say{true} values of the nation. Other highly salient populist issues, however, are almost impossible to express without recourse to nationalist ideology. Opposition to immigration and globalization, whether framed in economic or cultural terms, is fundamentally predicated on the drawing of boundaries around the nation, and these issues are often discussed in openly nationalist terms.

However, the affinity between nationalism and populism goes deeper than just these specific issue dimensions. Like populism, conceptualizing \say{the people} as a more or less unitary entity capable of expressing a will and in opposition to external enemies is central to nationalism as an idea \citep{anderson2006imagined}. The populist use of nationalism, therefore, relies on representing mainstream elites, past or present, as having failed to serve the nation's best interests in one way or another. This opposition also frequently takes on a territorial dimension, setting a spatially defined heartland that represents the true core of the nation against its domestic or foreign oppressors \citep{heinisch2019people}. In this regard, it is straightforward to see how lost territory can serve as a mobilizational resource for populists: not only is non-congruence of state with human boundaries the very definition of nationalism \citep{gellner2008nations}, but existing elites must also have failed to fulfill their patriotic duty by rectifying the situation. As such, among voters who identify strongly with the nation -- especially with a maximalist, trans-border conception thereof -- the issue of territorial loss is particularly well-suited to populist mobilization.

A second, almost universal feature of populist rhetoric that has rarely been remarked on is a tendency to focus on \textit{loss}. Whether in the economic or social domains, the emphasis is typically placed not on the possibility of improvement in the future, but on the harm that has already been done. In some cases, this loss may take the form of social status as a result of immigration and demographic shifts \citep{bustikova2014,mutz2018status}, while in others, it may be a loss of employment opportunities and relative income \citep{rodrik2018populism,rodrik2021does,rodriguez2021golfing}, but the representation of the present as in some way worse than the past is a constant feature of populist rhetoric. 

In addition to its natural affinity with the narrative representation of the people as in need of saving from harm, this tendency to present issues in terms of losses that have either been incurred or have begun to occur has a particular emotional power. The phenomenon of loss aversion is well-documented across a wide variety of domains and describes the general psychological tendency to respond to losses with strong negative emotions, such as anger and anxiety, that outweigh those experienced in response to an equivalent reward.\footnote{However, it is worth noting that this effect only persists as long as the loss stimulus is present, underling the importance of constant reinforcement of loss narratives to maintain salience.} The core prediction of prospect theory -- that this higher weighting of losses results in more acceptance of risk -- has been repeatedly confirmed in both behavioral and neurological experiments, although the exact mechanisms and role of emotional states remain unknown \citep{thaler1997effect,tom2007neural, sokol2019psychological,prietzel2020effect}. In the case of territory, the potential for loss framing is clear: to be considered lost, the land in question must necessarily have been part of the state at some point in the past (though in some cases, the link is quite tenuous). To the extent that this loss is then personally felt, it has the same power as issues such as immigration to induce the kinds of emotional states that tend to push voters toward populist parties \citep{zhou2023sources}. This effect may be particularly strong for newer populist parties perceived as political ``outsiders,'' given the element of risk in abandoning the political establishment for unproven newcomers.

The affinity between nationalist populism and the issue of \textit{territorial} loss in particular also relates to the position of national ``heartlands'' in the narratives these parties employ. \posscite{taggart2004populism} influential characterization of populism centers this element, arguing that the defining feature of populist rhetoric is its reliance on a reference point that evokes the idealized representatives of ``the people''; a ``middle England'' or ``middle America''. While this heartland need not be a real time or place at all, it tends to hearken back to an imagined golden age that evokes feelings of collective nostalgia \citep{lammers2020make}. This is precisely the same rhetorical strategy that undergirds irredentist nationalism, however: the desire to recover a lost homeland \citep{shelef2020homelands} or golden age \citep{cederman2024future} that is not only identified with control over the territory itself, but more broadly with the better times and sense of wholeness that it is associated with.

An important caveat is that, for efforts to prime voters with territorial losses to have an effect, those losses must resonate to some degree with voters. While the reasons some people place more value on lost territories than others fall outside of the scope of this paper, it is unlikely that reminding voters of losses will affect their behavior if the loss is viewed as inconsequential. At the same time, priming may also have little effect among the most committed irredentists if the consideration of past losses is already foremost in their minds. For this reason, we expect the effects of priming loss to be greatest among those with an intermediate level of concern.

In addition to representing current issues as losses, populists use narratives that embed these losses in a broader trend of \textit{victimization} by outside forces, activating status concerns and powerful emotions \citep{obradovic2020understanding}. While populists generally either blame domestic elites for the losses, or for having failed to reverse them, the nature of this blame need not necessarily go beyond corruption or incompetence. In the case of immigration, for example, it is generally difficult for all but the most determined conspiracy theorists to frame the issue as part of a concerted effort to harm \say{the people.} With other issues, such as opposition to international integration, however, it is much more straightforward to identify clear external culprits with hostile intent. Indeed, in Europe, it is precisely those nationalist populist parties that have been most successful in associating contemporary European Union policy disputes with historical memories of victimization that have enjoyed the greatest electoral success \citep{caramani2019national}. 

The potential to interpret territorial losses as instances of victimization is somewhat dependent on the circumstances of the loss in question. In cases of decolonization, for example, there is less scope for framing the metropole as a victim than when the loss occurred due to foreign conquest. However, in most cases, there is enough historical ambiguity that a loss can be readily blamed on foreign actors \citep{vardy1983impact,keyman2011tale,akccali2012geographical,narvselius2014lviv}. Although feeling victimized does not necessarily lead someone to vote for populists, the promise of \say{taking back control} and exercising agency that typically accompanies populist rhetoric is nonetheless particularly appealing in this context.

Finally, a corollary of focusing on losses is that populist issue framing is necessarily \textit{backward-looking}; it tends to emphasize a reference point in the national past that compares favorably to the present on the relevant dimension. For instance, politicians attempting to mobilize anxiety over value change might invoke memories (real or imagined) of an earlier time when traditional values predominated, while opponents of globalization might emphasize an earlier period when the nation enjoyed greater autonomy over its affairs. This association between the desired outcome and a point in the \textit{past} simultaneously delegitimizes \textit{current} elites, who are supposedly responsible for the nation's decline, while associating the populists with the certainty of the past. This latter point is particularly important for political outsiders, who may otherwise suffer from voter uncertainty over how they might behave if elected.

Territorial losses are, again, particularly well-suited to this kind of mobilization strategy because they are associated with a specific historical juncture. By associating the loss of territory not only with a violation of the principles of nationalism but also with a host of other ills that befell the nation in the intervening period, narratives of territorial loss can associate support for the populists with the prospect of returning to an idealized past. Indeed, this past need not even be within living memory: the memory of the Reconquista has been invoked by the VOX party in Spain to great effect, for example, despite having ended in the 15th century \citep{esteve2022reconquest}.  




We can, therefore, summarize the main observable implications of our theory as follows:

\begin{quote}
   \textsc{\textbf{Hypothesis 1}}\\
    The effect of priming territorial losses is to increase support for nationalist populist parties. 
\end{quote}

\begin{quote}
   \textsc{\textbf{Hypothesis 2}}\\
    Higher degrees of individual concern over lost territory lead to a greater likelihood of supporting nationalist populist parties.
\end{quote}
% \begin{enumerate}[label=\subscript{\textbf{H}}{{\arabic*}}: ]



%     \end{enumerate}
Although these features suggest a natural affinity between the mobilization of territorial loss and nationalist populist parties, it could, in principle, also be used by other parties. However, an important reason this is unlikely to be the case is the difficulty of building an actionable policy program. While simply lamenting the loss of territory may have some emotional resonance with voters, raising the salience of the issue will tend to create expectations that something will be done about it. With rare exceptions,\footnote{Recent notable instances of military solutions to the problem include Azerbaijan's 2023 reconquest of Nagorno-Karabakh and Russia's 2014 and 2022 invasions of Ukraine.} however, the options available to incumbents are extremely limited, typically not extending beyond the financing of cultural programs in the territory in question or the extension of citizenship to its residents. 

This relates closely to the existing debate over whether populist party platforms are compatible with governance, or whether being elected causes populists to moderate and become indistinguishable from their mainstream counterparts. While no clear conclusion has been reached, the weight of evidence suggests that, although populists are able to put much of their policy program into effect with measurable economic consequences, the substantive effect is largest for core policy areas such as immigration that are amenable to executive action \citep{albertazzi2015populists,pappas2019populists,zellman2019hawking,aytacc2021partisanship,kaltwasser2022populist,guriev2022political,funke2023populist}
. With regard to the foreign policy domain in particular, changes as result of electing populist parties have tended to be confined to \textit{tone} and \textit{style} rather than substance \citep{jenne2021populism,caiani2022three}. This reflects the greater constraint experienced by populists in power in this domain. Populists can change how they talk to or about international actors, but altering the substance, particularly regarding an issue as intractable as territorial control, is far more challenging.

As a result, parties with a history of governing are unlikely to make significant electoral gains from politicizing lost territory, as they must somehow account for their proven inability to resolve the issue. Certainly, this does not necessarily prevent populists \textit{in power} from doing so -- Viktor Orban in Hungary and Vladimir Putin in Russia have made extensive reference to lost territories while in office, for example, but their ability to continue mobilizing support over the issue is likely to be circumscribed by their ability to produce visible results.\footnote{Again, exceptions include Russia's annexation of Ukrainian territories (and, to a lesser extent, Turkey's control over parts of Syria), but these are  rare in recent history.}  Populist parties in opposition, by contrast, do not face this limitation and are consequently far more likely to benefit from raising the salience of territorial loss:

\begin{quote}
   \textsc{\textbf{Hypothesis 3a}}\\
    The effects of priming territorial loss concern on party approval are more positive for nationalist populists in opposition than for those in government.
\end{quote}
\begin{quote}
   \textsc{\textbf{Hypothesis 3b}}\\
    The effects of higher territorial loss concern on party approval are more positive for nationalist populists in opposition than for those in government.

\end{quote}


While populists may be able to raise the salience of territorial loss and affect its relationship to political decisions through framing, it is unlikely that they can shape actual attitudes, at least over the short term. Instead, we argue that individual attitudes toward lost territory are more or less stable over time and are shaped largely by the degree of identification with a spatially maximalist concept of the nation. This means that rather than being a byproduct of partisan preferences, we expect territorial loss concerns to be drivers of electoral choices.

\begin{quote}
   \textsc{\textbf{Hypothesis 4}}\\
    Attitudes to lost territory are stable over time and cause electoral behavior, and not the reverse.

\end{quote}

If territorial loss concerns are relatively fixed in the short term, this means that the electoral payoffs of playing the territorial loss card will depend on the prior distribution of territorial loss concerns among voters: for individuals who are bothered by past territorial losses, raising the salience of such losses should yield a greater boost in populist party support than for those who are less concerned about the lost territory. For the latter group, raising the salience of territorial losses may even backfire, as the specter of irredentist policies could raise concerns about their domestic opportunity costs and international risks. At the same time, the most committed irredentists are unlikely to be subject to priming effects since the issue of territorial loss is already highly salient. For this reason, we expect priming effects of the greatest magnitude among voters with high, but not extreme, levels of concern.

\begin{quote}
   \textsc{\textbf{Hypothesis 5}}\\
    The effect of priming concern on nationalist populist party approval is more positive among those who are more concerned about the loss, especially at intermediate levels.

\end{quote}
% \begin{quote}
%    \textsc{\textbf{Hypothesis 6}}\\
%     Priming territorial loss does not only affect political preferences, but also increases broader historical nostalgia for periods of greater territorial extent, even if non-democratic.

% \end{quote}

% \textbf{Hypothesis 6}: 


\section{Case Selection and Data}
    \subsection{Data}
    In order to test the theory outlined in the previous section, we rely on an original survey conducted in June-July 2020 in Germany, Romania, Hungary and Turkey with approximately 2,000 respondents in each country, hereafter referred to as the \say{loss survey} (LS). In addition to the geographic location of the respondents, the survey included a standard set of demographic measures, items measuring attitudes and beliefs regarding the COVID-19 pandemic, partisanship and voting questions, and a question order experiment manipulating the priming of territorial loss. 

The LS was fielded as an online survey by Qualtrics, which contracts with local panel providers. Respondents are recruited through a variety of non-monetary incentives and commit to complete a number of surveys while being in the panel. Survey invitations are sent to respondents on the basis of age, gender, education and region quotas, and while the resulting samples are still younger and more educated than the average population, the samples nevertheless capture a broad cross-section of the population of the four countries\footnote{We employ inverse probability sampling weights based on available demographic information in order to correct these imbalances.}.     
The experimental design, details of which can be found in Appendix E, consisted of independently randomizing the position of two items to be either before or after the political outcome battery, one of which measured concern over lost territory and the other support for a range of policies toward the foreign diaspora in each country except Germany. This design thus ensures that we have unbiased measures for the level of concern of each respondent, while also being able to estimate priming effects.  
In addition, follow-up surveys using the same methodology were fielded in Romania in December 2020-January 2021 and again in January 2022. In the 2021 wave, we asked respondents to report their vote choice in the December 2020 parliamentary elections, in which the nationalist populist AUR party, which was virtually unknown at the time of the first wave, received over 9\% of the  vote. Finally, in the most recent wave, the same question on concern for territorial loss was again included, allowing us to test the stability of attitudes over time. 
   
     Since no one commonly agreed-on classification exists with full coverage of these cases, we rely on a combination of multiple data sources to classify a party as nationalist populist. Specifically, we use a combination of the Chapel Hill Expert Survey (CHES) 2019 \citep{jolly2022chapel}, the PopuList dataset, \citep{rooduijn2024populist}, the Global Party Survey (GPS) \citep{norris2020measuring}, and the Varieties of party identity and organization (V-Party) dataset \citep{luhrmann2020varieties}, wherever available for a given party.  

     In order to evaluate whether a party is \textit{populist}, we average (after rescaling to ensure equal weight) the CHES ratings of position and salience of the people vs. elite dimension, the PopuList classification as populist, the analogous GPS classification, and the V-Party populism scale. An analogous procedure was performed to separately identify \textit{nationalist} parties, using position on minorities and nationalism from CHES and the equivalent scales for minority position and ``cultural superiority'' from V-Party. Parties that are above 0.5 on either dimension are classified as populist or nationalist, respectively, and those that meet both criteria are nationalist populist.

     Notably, while seven parties in our study (AKP and MHP in Turkey; PRM and PMP in Romania; Fidesz, and Jobbik in Hungary; and AfD in Germany) are classified as nationalist populist, only a single party (\textit{Die Linke}) emerges from this classification procedure as populist but not nationalist, while three (iYi in Turkey, PSD in Romania, and the CSU in Germany) are nationalist but not populist. \footnote{It is also worth noting that the nationalism score of these more mainstream nationalist parties were noticeably lower than those of their nationalist populist counterparts}   While we address the question of whether our results are driven by populism or nationalism briefly in a later section and more extensively in Appendix A, this strong observed association between populism and nationalism both lends weight to our theoretical argument and makes it much harder to disentangle the two empirically.
    
   % We first report the effect of priming territorial loss on ratings of parties that are identifiably nationalist populist based on the Chapel Hill Expert Survey (CHES) 2019.\footnote{Specifically, we code a party as nationalist populist if it scores above the sample median for 2019 on at least 2 of: GALTAN, People vs. Elite position and Nationalism. For our cases, this yields a classification of AfD (Germany), Fidesz (Hungary), Jobbik (Hungary), AKP (Turkey), MHP (Turkey), IYI (Turkey), PRM (Romania) and PMP (Romania) as nationalist populist.} 
   
   
    \par
    \subsection{Context and Background}
    The selection of Romania, Hungary, Turkey and Germany as cases\footnote{A reduced version of the same study was also fielded in Russia and yielded similar results, although we do not focus on this case in the present paper due to the non-competitive political environment.} provides variation on four major dimensions of interest: the political salience of lost territory, how and when the territory was lost, whether it is now inhabited by co-nationals and the political meaning of the loss, summarized in Table \ref{tab:cases}. While the salience of the respective losses has been extremely high in all four cases at some point in the past, at the time of the present study it is a near-constant feature of political discourse in Hungary and, to a lesser extent, Romania and Turkey. In Germany, by contrast, discussion of lost territories is largely absent from mainstream political narratives. These differences are also reflected in the survey-based levels of average individual salience, which were highest in Hungary, closedly followed by Romania and Turkey, with Germany significantly lower. \par
    These cases also vary considerably regarding the circumstances of the loss. While Hungary, Romania, Turkey and Germany all experienced their losses in approximately the same period -- during the first half of the 20th century as a result of the two world wars -- the magnitude and type of each loss is distinct. Whereas in the aftermath of World War I, Hungary lost approximately two-thirds of its territory, much of which continues to be inhabited by Hungarian-speakers, to neighboring countries, Romania's 1940 losses at the hands of the Soviet Union were relatively smaller, consisting primarily of Moldova, as well as parts of northern Bulgaria and Ukraine. These losses were accompanied by forced resettlements so only Moldova and northern Bukovina remain occupied by \say{co-nationals,} although the status of the Moldovan language and its speakers as Romanians remains contested in Moldova.
    \par
    Germany's losses consisted primarily of parts of present-day Poland, France, and the Czech Republic, which were lost after both world wars and (with the exception of the Alsace-Lorraine area in France) are no longer inhabited by significant numbers of German speakers due to large-scale evictions at the end of WWII. However, the specific circumstances of Germany's involvement in World War II have given them a different character than in Hungary or Romania. Whereas the original losses following the First World War were readily interpreted as unfair victimization by foreign powers, akin to those of Hungary or Romania, Germany's role as an instigator of WWII -- and the prominence of irredentism and expansionism in the Nazi political project -- has largely reversed this narrative and tied the territorial loss conceptually to the horrors of the Nazi era.
    \par
    Turkey's losses represent yet another model. While contiguous, they represent not the loss of \say{core} Turkish territory but rather the loss of a multinational empire following the Ottoman defeat in the First World War. Turkey thus stands out from the other cases not only in having lost a vastly greater amount of territory\footnote{As one of the Turkish survey respondents noted, Turkey had lost ``Europe, Asia and Africa''} but also in that very little of that territory is easily constructed as part of the Turkish nation without conflating nation with empire. It is also the only country in the sample that has engaged in military action in its former territories in the post-war period -- albeit not with irredentist aims-- including intervention in Cyprus in 1974 and the recent incursion into Northern Syria.

    \begin{table}[!htbp] \centering
\caption{Summary of Cases}
\label{tab:cases}
\resizebox{\textwidth}{!}{
    \setlength{\extrarowheight}{2pt}
    \begin{tabular}{cc|c|c|c|c|}
      & \multicolumn{1}{c}{} & \multicolumn{4}{c}{Case}\\
      & \multicolumn{1}{c}{}  & \multicolumn{1}{c}{\textit{Romania}} & \multicolumn{1}{c}{\textit{Hungary}} & \multicolumn{1}{c}{\textit{Turkey}} & \multicolumn{1}{c}{\textit{Germany}} \\\cline{3-6}
      \multirow{21}*{Dimension}  & \multirow{4}*{\textit{Main Losses}}   &  & & &\\
      & & Bessarabia (Moldova) & Transylvania, Vojvodina & Former Ottoman & Silesia/East Prussia, Sudetenland\\
      &   & South Dobruja, Bukovina & Upper Hungary & Empire & Alsace-Lorraine\\
      &   &  & & &\\\cline{3-6}
      & \multirow{2}*{\textit{Average concern 0--100}}  & \multirow{2}*{58} & \multirow{2}*{61} & \multirow{2}*{58} & \multirow{2}*{19}\\
      &   &  & & &\\\cline{3-6}
      & \multirow{4}*{\textit{Circumstances of Loss}}   &  & & &\\
      &  & 1940, 1944 & 1920 & 1918 & 1918, 1945\\
      &   & Soviet annexation & Defeat in WWI & Defeat in WWI & Defeat in WWI/II\\
      &   &  & & &\\\cline{3-6}
      & \multirow{2}*{\textit{Co-Nationals}} & \multirow{2}*{Numerous} & \multirow{2}*{Numerous} & \multirow{2}*{Few\tablefootnote{While most of the territories comprising the former Ottoman Empire contain, at most, small populations of ethnic Turks, adherents of Pan-Turkist ideology adopt a far more expansive definition of the nation, which notably includes Azerbaijan.}} & \multirow{2}*{Relatively few}\\
      &   &  & & &\\\cline{3-6}
      & \multirow{6}*{\textit{Nationalist Populist Parties}} &  &  & & \\
      &   &  & & &\\
      & & PRM (Opposition), PMP (Opposition)  & Fidesz (Governing) & AKP (Governing) & AfD (Opposition)\\
      &  &  & & &\\
      &  & AUR (Opposition, new in 2020) & Jobbik (Opposition) & MHP (Governing)\  & \\
      &   &  & & &\\
      &   &  & & &\\\cline{3-6}
    \end{tabular}
}
\vspace{\fill}
\end{table}


\section{Research Design}

    The experimental component of our analysis is based on a question order experiment: all participants were asked the same question but at different locations within the survey. Thus, in one version of the survey, respondents are asked about lost territory before answering outcome items, while, in the other the outcome measures precede the lost territory prime (see Table A.8 for details of the question order). A third version of the survey includes a question about policy preferences toward diaspora populations prior to the outcomes. We exclude all subjects assigned to this condition from the main experimental analyses due to the possibility of cross-priming, although the main results are unaffected by this decision (see the online appendix). %While in principle this generates a factorial design, the two questions prime essentially the same issue, so we report main results treating the presence of \textit{either} as priming lost territory\footnote{All main results are robust to alternative specifications.}.  \par 

    The question serves as both an informational prime and a data source. The territorial loss treatment begins by providing a brief, neutrally worded informational statement about the loss of territory before asking respondents to recall a specific territory and report how concerned they are about its loss. 
    \begin{quote}
    ``There have been a number of dramatic changes to Europe’s borders in the past century, and
much land that belonged to [Country] in previous generations has now been lost to other
countries. Please use the field below to enter the first such territory that comes to mind,
or, if you cannot remember its name, then enter the country to which it now belongs. [Text
field] [New Page] Please indicate on a scale from 0 to 100, where 100 is ‘very concerned’ and
0 is ‘not at all concerned’, how concerned you are about the loss of this territory.''
\end{quote}
Analogous to \say{racial priming} experiments in research on race in American politics \citep{mendelberg2017race}, this is thus a minimal treatment, intended not to persuade respondents or to elicit an emotional response but simply to \textit{prime} the issue of territorial loss by asking them to actively recall a specific example. Therefore, it should not be expected to make respondents significantly \textit{more concerned} about loss of territory but rather to increase the salience of pre-existing attitudes with respect to other choices. Moreover, multiple intervening questions between the prime and the main outcomes makes this a hard test of priming effects, as they must persist over time.  \par 
    %Similarly, the diaspora treatment simply asks respondents to express preferences on a number of policies, arranged by increasing risk (from funding cultural programs to military posturing), with only a minimal prime that \say{many people are concerned about the way they [co-nationals] are being treated}. This item thus serves as both treatment and outcome; it provides a direct measure of risky policy preference on an issue for which territorial loss is naturally salient, but its connection with the issue of territorial loss is sufficiently strong that it can plausibly be interpreted as priming similar, if not the same, concerns.
    \par
    Additionally, since the loss treatment is expected to prime territorial loss, we are interested in both the direct effect of the treatment and its \textit{interaction} with reported concern over loss. If the theory is correct, then the effect of the treatment would be to increase the effect of higher levels of concern on other decisions. While this measure is available for all respondents regardless of question order, it is inherently a post-treatment measure since the item measuring reported concern also primes it. For this reason, we exclude reported concern from the experimental control variables except when explicitly considering heterogeneous effects. While baseline (pre-prime) concern is inherently unobservable, we do not expect that this measure would substantially bias results since there is no theoretical reason to suppose that the question order manipulation would systematically affect respondents' reported concern.\footnote{Indeed, there is no statistically significant difference in reported concern across experimental conditions in the data.}
     \par
     In addition to these two primary measures, we include a set of standard socio-demographic control variables in the baseline specification, including age, education, gender and settlement size. Since we are concerned with the role of territorial loss as an independent predictor of nationalist populist support, we additionally include a set of indices intended to capture the main attitudinal drivers of populism suggested by the literature: religiosity, political interest, LGBTQ+ attitudes, national identity, attitudes to democracy, support for globalization, opposition to COVID-19 restrictions, and support for interventionist economic policy. More details on the construction of these indices can be found in Appendix D. 
     
     Missingness overall is extremely low across the board, with fewer than 0.1\% of respondents across all cases missing responses to either the territorial loss item or our party rating outcomes\footnote{The major exception to this is the variable indicating whether respondents voted for a nationalist populist candidate at the previous election, which is missing for 17\% of respondents. However, this is not used in any of the main analyses.}. The multi-item nature of these indices, however, results in slightly higher levels of missingness, with the itemwise maximum being 4\% of respondents for whom the globalization support index is unavailable. To mitigate this missingness we report our main results based on five multiple imputations using classification and regression trees (CART); however, all results are substantively unaffected by this decision. 
     
    Finally, the panel structure of the data makes it possible to assess the actual behavioral consequences of holding attitudes toward lost territory, mitigating the typical difficulty of connecting hypothetical choices in survey experiments to real outcomes of interest, as well as overcoming the problem of measuring loss attitudes without simultaneously priming them. Between the first and third waves of the panel survey, a parliamentary election was held in Romania (Dec. 6, 2020), in which a new nationalist populist party, AUR, achieved an unexpectedly high share, becoming the second largest opposition party with just over 9\% of the vote. Although the party officially formed in 2019, its rise to national prominence occurred after the first wave of the panel survey, due to a combination of criticism of government-enacted COVID-19 policies and a platform that emphasized the need to push for reunification with Moldova.\footnote{As late as October 2020, two months before the parliamentary elections, AUR only received .33\% of the vote in the local elections.}
    \par
    By examining the effect of attitudes towards territorial loss held in the first wave on subsequent vote choice, we can estimate the effect of holding such attitudes on voting for populists without concern for the possibility of reverse causality. Since respondents were unlikely to have been exposed to any AUR messaging at the time of the initial wave, we can be confident that their expressed level of concern was not a result of partisan-motivated reasoning or of opinion leadership from the party elite. While this does not definitively establish causality, as it is still possible that an unmeasured third factor causes both vote choice and attitudes, we can minimize this concern by controlling for other factors that might plausibly influence vote choice.


     \subsection{Attitudes to Territorial Loss}
    \label{sec:maps}
     Before presenting the main experimental results, we provide a descriptive overview of the general attitudinal patterns observed in the data. While all findings presented in this section are necessarily correlational, to our knowledge they represent the first systematic comparative evidence on the prevalence of popular irredentism in countries without active territorial conflicts.

    In all cases except Germany, lost territory has a high baseline level of salience, as evidenced by Figure \ref{fig:concerndist}, with responses in the 95-100 range as the modal category in Romania, Hungary and Turkey for an item tasking respondents with indicating their level of concern on a 100-point \say{feeling thermometer} scale. Although all samples display clustering at the endpoints and midpoints, the item has support throughout its entire range, suggesting that the feeling thermometer captures meaningful variation in attitudes at all levels. Notably, even in Germany, where the largest number of respondents reported 0 concern over lost territories, an absolute majority (60\%) nevertheless indicated a positive value in response to the item, reflecting the enduring resonance of this issue. \par
\begin{figure}
    
    \centering
    \makebox[\linewidth][c]{\begin{subfigure}[t]{0.45\textwidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/romdist.png} 
        \caption{Romania} \label{fig:distrom}
    \end{subfigure}
    \begin{subfigure}[t]{0.45\linewidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/hundist.png} 
        \caption{Hungary} \label{fig:disthun}
    \end{subfigure}}

    \vspace{1cm}
    \makebox[\linewidth][c]{\begin{subfigure}[t]{0.45\linewidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/turdist.png} 
        \caption{Turkey} \label{fig:disttur}
    \end{subfigure}
    \begin{subfigure}[t]{0.45\linewidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/gerdist.png} 
        \caption{Germany} \label{fig:distger}
    \end{subfigure}}
    \caption{Distribution of concern about lost territories on a 100-point feeling thermometer scale by country.}
    \label{fig:concerndist}
\end{figure}

     Table \ref{tab:concernpred} gives each country's major demographic and attitudinal correlates of loss concern based on a simple multiple regression. The dependent variable in these regressions was the level of concern for lost territories (measured on a 0-100 scale), while the predictors included the level of identification with each country's culture, nation and state (measured on a 5-point scale), along with a series of demographic indicators and partisanship. The results in Table \ref{tab:concernpred} reflect important differences in the meaning of lost territory in each country. In Romania and Hungary, where higher education often contains a significant patriotic component, frames of national victimization vis-à-vis the territorial losses are widespread and significant diasporas inhabit the lost territories; education and identification with the nation, state and ethnic groups are strongly associated with concern over lost territory. In Turkey and Germany, by contrast, higher education is weakly negatively associated with average concern over territory, although the relationship is not significant. Notably, support for democracy is also negatively associated with concern over lost territory in Hungary and Germany, where irredentism is strongly associated with anti-democratic political actors. These differences across cases are summarized in Table \ref{tab:cases}.
\begin{table}[!htbp] \centering 
  \caption{Correlates of Concern over Territorial Loss by Country} 
  \label{tab:concernpred} 
\begin{adjustbox}{max width=\textwidth}
\begin{tabular}{@{\extracolsep{5pt}}lccccc} 
\\[-1.8ex]\hline \\[-1.8ex] 
\\[-1.8ex] & \multicolumn{5}{c}{Territorial Loss Concern} \\ 
 & Hungary & Romania & Turkey & Germany & Pooled \\ 
\\[-1.8ex] & (1) & (2) & (3) & (4) & (5)\\ 
\hline \\[-1.8ex] 
 Age & $-$0.088 & 0.077 & $-$0.050 & $-$0.071 & $-$0.025 \\ 
  & (0.064) & (0.079) & (0.126) & (0.056) & (0.041) \\ 
  Completed University & 6.001$^{*}$ & 8.629$^{**}$ & 0.980 & 0.493 & 4.472$^{**}$ \\ 
  & (3.277) & (3.986) & (3.845) & (3.073) & (2.188) \\ 
  Completed Secondary & 7.786$^{**}$ & 6.060 & $-$0.178 & $-$1.864 & 4.626$^{*}$ \\ 
  & (3.380) & (4.028) & (4.637) & (3.425) & (2.499) \\ 
  Technical/Vocational Education & 4.316 & 5.971 & 1.112 & 4.709 & 4.600$^{**}$ \\ 
  & (3.097) & (5.676) & (4.026) & (2.940) & (2.303) \\ 
  Male & 7.815$^{***}$ & 5.949$^{**}$ & $-$6.859$^{**}$ & 0.596 & 4.095$^{***}$ \\ 
  & (1.801) & (2.644) & (3.011) & (2.129) & (1.364) \\ 
  National Language Speaker & 2.626 & 4.423 & 7.402 & $-$0.607 & 2.905 \\ 
  & (3.376) & (4.684) & (4.911) & (3.297) & (2.443) \\ 
  COVID Restriction Support Index & 4.291 & 5.693 & 6.624 & 9.622$^{**}$ & 6.914$^{**}$ \\ 
  & (4.294) & (5.457) & (7.252) & (4.547) & (2.920) \\ 
  Interventionism Index & 10.120$^{**}$ & 3.494 & 24.245$^{***}$ & $-$0.748 & 9.243$^{***}$ \\ 
  & (4.771) & (4.536) & (7.325) & (5.172) & (3.368) \\ 
  Globalization Support Index & $-$4.073 & $-$3.110 & $-$26.095$^{***}$ & $-$11.729$^{**}$ & $-$14.327$^{***}$ \\ 
  & (4.714) & (6.256) & (6.542) & (5.231) & (3.129) \\ 
  Support for Democracy Index & $-$9.010$^{**}$ & $-$2.852 & $-$21.871$^{***}$ & $-$21.589$^{***}$ & $-$10.973$^{***}$ \\ 
  & (4.352) & (7.448) & (7.062) & (7.297) & (3.468) \\ 
  Accept LGBT Neighbors & $-$7.106$^{**}$ & $-$3.457 & 2.419 & $-$20.387$^{***}$ & $-$1.691 \\ 
  & (3.119) & (3.242) & (4.264) & (4.879) & (2.339) \\ 
  Religious Importance & 9.638$^{***}$ & 5.961 & $-$2.767 & 17.812$^{***}$ & 8.860$^{***}$ \\ 
  & (2.894) & (4.514) & (5.688) & (3.599) & (2.249) \\ 
  Political Interest & $-$2.430 & $-$5.047 & $-$3.088 & $-$6.642$^{*}$ & $-$8.605$^{***}$ \\ 
  & (3.288) & (4.656) & (6.102) & (4.010) & (2.570) \\ 
  Cultural ID & 3.825$^{***}$ & 6.681$^{***}$ & 4.925$^{**}$ & 2.224 & 3.602$^{***}$ \\ 
  & (1.412) & (2.346) & (2.327) & (1.548) & (1.048) \\ 
  National ID & 4.754$^{***}$ & 2.705 & 2.023 & $-$0.767 & 2.576$^{**}$ \\ 
  & (1.610) & (2.372) & (2.161) & (1.791) & (1.082) \\ 
  State ID & 3.262$^{***}$ & 1.171 & $-$4.286$^{**}$ & 0.991 & 0.580 \\ 
  & (1.228) & (1.810) & (1.944) & (1.554) & (0.975) \\ 
  Urban & 5.031$^{***}$ & 0.015 & $-$1.498 & $-$1.160 & 1.070 \\ 
  & (1.936) & (2.584) & (3.463) & (1.922) & (1.195) \\ 
 N & 1,445 & 1,430 & 1,161 & 1,400 & 5,463 \\ 
\hline \\[-1.8ex] 
\multicolumn{6}{l}{$^{*}$p $<$ .1; $^{**}$p $<$ .05; $^{***}$p $<$ .01} \\ 
\multicolumn{6}{l}{Region fixed effects included for all models} \\ 
\end{tabular} 
\end{adjustbox}
\end{table} 
%     \begin{table}[!htbp] \centering 
%   \caption{Correlates of Concern over Territorial Loss by Country} 
%   \label{tab:concernpred} 
% \begin{adjustbox}{max width=\textwidth}
% \begin{tabular}{@{\extracolsep{5pt}}lccccc} 
% \\[-1.8ex]\hline \\[-1.8ex] 
% \\[-1.8ex] & \multicolumn{5}{c}{Territorial Loss Concern} \\ 
%  & Hungary & Romania & Turkey & Germany & Pooled \\ 
% \\[-1.8ex] & (1) & (2) & (3) & (4) & (5)\\ 
% \hline \\[-1.8ex] 
%  Age & $-$0.098 & 0.030 & $-$0.136 & $-$0.052 & $-$0.058 \\ 
%   & (0.059) & (0.081) & (0.123) & (0.058) & (0.036) \\ 
%   Completed University & 5.389$^{*}$ & 9.101$^{**}$ & $-$1.068 & $-$2.397 & 2.246 \\ 
%   & (3.018) & (3.881) & (3.751) & (3.205) & (1.810) \\ 
%   Completed Secondary & 8.760$^{***}$ & 6.510$^{*}$ & $-$1.529 & $-$3.144 & 3.721$^{*}$ \\ 
%   & (3.127) & (3.832) & (4.464) & (3.518) & (2.046) \\ 
%   Technical/Vocational Education & 6.099$^{**}$ & 8.000 & $-$0.166 & 4.156 & 4.344$^{**}$ \\ 
%   & (2.822) & (5.332) & (4.017) & (3.054) & (1.842) \\ 
%   Male & 7.666$^{***}$ & 5.931$^{**}$ & $-$6.466$^{**}$ & 1.791 & 3.003$^{***}$ \\ 
%   & (1.619) & (2.496) & (2.987) & (2.154) & (1.106) \\ 
%   National Language Speaker & 2.273 & 3.626 & 7.226 & $-$5.104 & 0.277 \\ 
%   & (3.234) & (4.839) & (4.600) & (3.472) & (2.011) \\ 
%   Political Interest & 0.855 & $-$0.064 & 2.181 & 1.734 & 1.673$^{***}$ \\ 
%   & (0.788) & (1.226) & (1.451) & (1.077) & (0.543) \\ 
%   Cultural ID & 3.825$^{***}$ & 7.404$^{***}$ & 5.802$^{**}$ & 2.753$^{*}$ & 4.100$^{***}$ \\ 
%   & (1.306) & (2.260) & (2.251) & (1.559) & (0.916) \\ 
%   National ID & 6.114$^{***}$ & 1.507 & 0.220 & $-$0.978 & 2.285$^{**}$ \\ 
%   & (1.489) & (2.318) & (2.143) & (1.791) & (0.992) \\ 
%   State ID & 2.992$^{***}$ & 3.561$^{**}$ & $-$3.705$^{*}$ & 1.276 & 1.885$^{**}$ \\ 
%   & (1.097) & (1.795) & (2.031) & (1.518) & (0.780) \\ 
%   Rural & $-$5.459$^{***}$ & 0.440 & $-$7.482 & 0.858 & $-$2.266 \\ 
%   & (2.041) & (2.982) & (6.712) & (2.949) & (1.547) \\ 
%   Urban & $-$4.852$^{*}$ & 0.700 & $-$0.978 & 1.834 & $-$0.517 \\ 
%   & (2.477) & (2.625) & (3.385) & (2.493) & (1.413) \\ 
%   Support for Democracy & $-$3.325$^{***}$ & $-$2.845 & $-$3.221$^{*}$ & $-$6.776$^{***}$ & $-$3.569$^{***}$ \\ 
%   & (0.966) & (1.743) & (1.844) & (1.763) & (0.704) \\ 
%  N & 1,739 & 1,629 & 1,315 & 1,530 & 6,249 \\ 
% \hline \\[-1.8ex] 
% \textit{Note: Region fixed effects included for all models}  & \multicolumn{5}{r}{$^{*}$p$<$0.1; $^{**}$p$<$0.05; $^{***}$p$<$0.01} \\ 
% \end{tabular} 
% \end{adjustbox}
% \end{table} 
In addition to numerical ratings of concern, the prime required respondents to name a specific territory that first came to mind. Table \ref{tab:salienceprops} gives the overall distribution of territories named for the whole sample in each case. Notably, although a single territory is noticeably more salient than others in all four cases, only in Romania and Hungary do a majority of respondents name a single territory. In all four cases, moreover, at least four distinct territories are named by over 5\% of respondents, underlining the rich array of potential irredentist issues available to political entrepreneurs.

\begin{table}[!htbp]
\centering 
  \caption{Proportion of Respondents Naming Most Salient Territories by Country} 
  \label{tab:salienceprops} 
  \begin{adjustbox}{max width=0.8\textwidth}
\begin{tabular}{ll}
  \hline
Territory & Proportion\tablefootnote{Note that some respondents listed more than one territory without prompting, so percentages do not necessarily total 100\%.} \\ 
  \hline
\multicolumn{2}{l}{\textbf{Hungary}}\\
\vspace{0.4em}\\
 Transylvania & 69.3 \% \\ 
  Upper Hungary (Slovakia) & 16.8 \% \\ 
  Kingdom of Hungary & 6.9 \% \\  
  Vojvodina (Serbia) & 6.1 \% \\
  Transcarpathia (Ukraine) & 5.6 \% \\  
  Other & 14.3 \% \\ 
  DK/RA & 2 \% \\ 
   \hline
\multicolumn{2}{l}{\textbf{Romania}}\\

\vspace{0.4em}\\
Bessarabia (Moldova) & 65.4 \% \\ 
  Bukovina (Ukraine)& 12.1 \% \\ 
  South Dobruja (Bulgaria) & 8.2 \% \\ 
  Other & 16 \% \\ 
  DK/RA & 3.8 \% \\ 
   \hline
\multicolumn{2}{l}{\textbf{Turkey}}\\

\vspace{0.4em}\\
Greece/Aegean Islands & 36.7 \% \\ 
  Iraq & 17.1 \% \\ 
  Syria & 17 \% \\ 
  Balkans & 12.9 \% \\ 
  Bulgaria & 11.6 \% \\
  Arabia/North Africa & 7.2 \% \\ 
  Ottoman Empire/Turkey & 6 \% \\ 
  Other & 16.8 \% \\ 
  DK/RA & 8.9 \% \\ 
   \hline
\multicolumn{2}{l}{\textbf{Germany}}\\  
\vspace{0.4em}\\
  Silesia/East Prussia (Poland) & 41.1 \% \\ 
  Alsace (France) & 12.1 \% \\
  Kaliningrad (Russia) & 10.7 \% \\ 
  Sudetenland (Czech Republic) & 5.1 \% \\  
  Other & 18.7 \% \\ 
    DK/RA & 14.8 \% \\ 
   \hline
\multicolumn{2}{l}{}\\
\end{tabular}
\end{adjustbox}
\end{table}

Attitudes toward lost territory also vary spatially within cases, summarized in Figure \ref{fig:concerndistgeo}, which displays average responses by County (NUTS level 3) for Romania and Hungary and by L\"{a}nder/Region (Nuts level 1) for Germany and Turkey, where fine-grained geographical information is not available. As might be expected, the general pattern is an increase in average concern near borders with lost territories. This is particularly prevalent in Romania, where the highest concern areas are those nearest Moldova. However, similar patterns can be observed in Germany (with Saarland and Baden-W\"{u}rttemberg displaying particularly high levels of concern) and in the easternmost regions of Turkey. However, this pattern does not hold for all country-territory pairs: in Hungary, for instance, concern is high in all regions, with only slightly higher average values in the regions bordering Romania and Ukraine. In contrast, the western regions of Turkey, despite being the most proximate to the high-salience Aegean Islands (see Table \ref{tab:salienceprops} for a full breakdown of named territories by case), display the lowest overall salience of lost territory.

The choice of territory is also affected by within-country geographic differences. Figure \ref{fig:saliencedistgeo} shows the distribution of respondents naming the overall most salient territory: Moldova/Bessarabia in Romania, Transylvania in Hungary, Greece (particularly the Aegean Islands) in Turkey and Poland/Silesia in Germany. In Romania, Turkey and Germany, the expected pattern of respondents living closer to the territory in question and naming it with higher frequency strongly holds. In Hungary, in contrast, Transylvania is mentioned by the highest proportion of respondents in the regions \textit{furthest} away from the Romanian border. 


\begin{figure}
    \centering
    \makebox[\linewidth][c]{\begin{subfigure}[t]{0.45\textwidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/romdistgeo.png} 
        \caption{Romania} \label{fig:distromgeo}
    \end{subfigure}
    \begin{subfigure}[t]{0.45\linewidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/hundistgeo.png} 
        \caption{Hungary} \label{fig:disthungeo}
    \end{subfigure}}

    \vspace{1cm}
    \makebox[\linewidth][c]{\begin{subfigure}[t]{0.45\linewidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/turdistgeo.png} 
        \caption{Turkey} \label{fig:distturgeo}
    \end{subfigure}
    \begin{subfigure}[t]{0.45\linewidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/gerdistgeo.png} 
        \caption{Germany} \label{fig:distgergeo}
    \end{subfigure}}

    \caption{Spatial distribution of concern about lost territories on a 100-point feeling thermometer scale by country.}
    \label{fig:concerndistgeo}
\end{figure}


\begin{figure}
    \centering
    \makebox[\linewidth][c]{\begin{subfigure}[t]{0.45\textwidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/rom_bessar.png} 
        \caption{Romania} \label{fig:rom_bessar}
    \end{subfigure}
    \begin{subfigure}[t]{0.45\linewidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/hun_erdely.png} 
        \caption{Hungary} \label{fig:hun_erdely}
    \end{subfigure}}

    \vspace{1cm}
    \makebox[\linewidth][c]{\begin{subfigure}[t]{0.45\linewidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/tur_greece.png} 
        \caption{Turkey} \label{fig:tur_greece}
    \end{subfigure}
    \begin{subfigure}[t]{0.45\linewidth}
        \centering
        \includegraphics[width=.95\textwidth]{Figures/ger_silesia.png} 
        \caption{Germany} \label{fig:ger_silesia}
    \end{subfigure}}    

    \caption{Spatial distribution of proportion of respondents reporting highest-salience lost territory.}
    \label{fig:saliencedistgeo}
\end{figure}


\section{Results}

    \subsection{Territorial Loss Attitudes and Nationalist Populist Party Support}
    \label{sec:expresults}
    
    We first report the effect of priming territorial loss on ratings of parties that are identifiably nationalist populist. Since complete randomization guarantees that the difference-of-means is an unbiased estimator for the average treatment effect (ATE), all main results come from bivariate weighted OLS. Since in general the population average treatment effect (PATE) 
    is different from the sample average (SATE) when treatment effects are not constant, our main findings are reported using inverse probability sampling weights. SATE estimates, which do not differ substantially, can be found in Appendix C, along with full control specifications to increase the efficiency of the estimator. For the observational results reported later, we additionally include a fixed set of controls, consisting of age, education, gender, urban status, political interest, religiosity, political interest, LGBTQ+ attitudes, national identity, attitudes to democracy, support for globalization, opposition to COVID-19 restrictions, and support for interventionist economic policy. The effects on the highest\footnote{We also repeat the analysis using average ratings.  While this choice does not dramatically affect results, the maximum specification is preferred since, in some instances populist parties exist on opposite ends of the political spectrum (e.g. in Germany), while in other cases nationalist populist parties are in opposing political camps (e.g. in Hungary), which raises the possibility that increases in ratings for some parties and a corresponding decrease in ratings of their opponents may therefore result in erroneous null estimates of average effects.} rating assigned to any of the nationalist populist parties in the pooled model are shown graphically in Figure \ref{fig:popratings}, as well as disaggregated by the parties in power (Fidesz in Hungary and the AKP and MHP\footnote{While the MHP were not represented in government at the time of the survey, we include them under this category, given their open support of Erdogan and the AKP. The results are unchanged by their inclusion or exclusion.} in Turkey) and those in opposition. Average treatment effect estimates of the prime are given in Figure \ref{fig:popratingexp}, while Figure \ref{fig:popratingobs} shows the estimated effects of reported concern. The predicted probabilities of retrospective vote choice, along with the full set of coefficient estimates across multiple specifications, can be found in the Appendix. \footnote{Since vote choice occurred prior to the survey and reports are thus unlikely to have been influenced by the prime, these effects should not be interpreted causally but provide additional evidence of an association between loss attitudes and behavior.}
    \begin{figure}[htp]
    \centering
    \makebox[\linewidth][c]{\begin{subfigure}[t]{0.45\textwidth}
        \centering
        \includegraphics[width=\textwidth,height=50mm]{Figures/Experimental Effects.png} 
        \caption{Average Treatment Effect Estimates of \\Experimental Prime} \label{fig:popratingexp}
    \end{subfigure}
    \begin{subfigure}[t]{0.45\textwidth}
        \centering
        \includegraphics[width=\textwidth,height=50mm]{Figures/Observational Effects.png} 
        \caption{Coefficient Estimates for Reported Concern} \label{fig:popratingobs}
    \end{subfigure}}
    \caption[Effects of Prime and Loss Concern on Nationalist Populist Party Ratings by Opposition Status]{Effects of Prime and Loss Concern on Nationalist Populist Party Ratings by Opposition Status\footnotemark}
    \label{fig:popratings}
    \end{figure}





    Overall, the results are strongly supportive of $\mathbf{H}_2$: As illustrated in Figure 4b, territorial\footnotetext{See Tables C1 and C2 in the Online Appendix for full model specifications.} loss concerns are strongly and robustly associated with nationalist populist party ratings, with a shift from 0 to 100 concern equating to an increase in average ratings of about 10 points on the 0-100 popularity scale, or approximately one-third of a standard deviation in the DV. Furthermore, the results in model 3 of Table A1 in the Appendix suggest that the territorial loss concern effect is similar in magnitude to the effects of prominent drivers of nationalist populist party support, such as nationalism and anti-LGBT attitudes, and more than three times larger than opposition to globalization. Meanwhile, the results in Figure 4a provide weaker support for $\mathbf{H}_1$. While priming respondents with territorial loss provides a slight boost to ratings of nationalist populist parties, the magnitude of the effect was rather small (8\% of a standard deviation) and fell short of achieving statistical significance at the 5\% threshold. This (non)finding  suggests that -- at least on average -- nationalist populist parties do not stand to gain much from simply raising the salience of territorial loss concerns, despite the robust association between attitudes and party preferences. 

    In order to test $\mathbf{H}_{3a}$ and $\mathbf{H}_{3b}$, we also disaggregate the outcome measure to  ratings of parties in power versus those in opposition, shown in the bottom part of Figure \ref{fig:popratings}. Since only two countries -- Hungary and Turkey -- had nationalist populist parties in power at the time of the surveys, the sample size is cut in half for these models, resulting in somewhat larger standard errors. The patterns suggest some tentative support for $\mathbf{H}_{3a}$, since the effects of the territorial loss prime were positive and marginally statistically significant for opposition nationalist populists but virtually zero for governing nationalist populists.\footnote{However, we should note that the difference between the two estimates fell short of achieving statistical significance at conventional levels, largely due to the large confidence intervals of estimates for governing nationalist populists.}   In Figure 4b, we find strong support for $\mathbf{H}_{3b}$: While the effects of territorial loss concerns on the popularity of governing nationalist populist parties are fairly small and statistically insignificant at conventional levels,  for opposition nationalist populists, the effects are highly significant and substantively large (about half a standard deviation of the DV).\footnote{Moreover, the difference in territorial loss concern coefficients between governing and opposition nationalist populists was statistically significant at .05).} 
    

    
    %Moreover, the association generally runs in the expected direction, with higher levels of concern associated with higher ratings for right-populist parties such as Jobbik and Fidesz in Hungary, the PRM in Romania, and the AfD in Germany. The priming effects, though largely non-significant, are also in the expected direction, with large negative effects on the centre-left USR and centre-right PNL in Romania and a large positive effect among respondents with high levels of concern on ratings for the opposition party IYI in Turkey, which has a history of politicizing border issues, especially regarding the Aegean islands. 
    

    % However, as is clear from Figures \ref{fig:popratingpower} and \ref{fig:popratingopp}, this relatively weak effect of the prime in general is largely explained by the opposite effects on attitudes toward populists in and out of government. Consistent with H3, priming respondents with lost territory significantly \textit{worsens} their attitude towards populists in power, especially for those with low baseline levels of concern. Indeed, only those with levels of concern above 75 have predicted party ratings greater than those for the control group. In contrast (Figure \ref{fig:popratingopp}), opposition parties enjoy substantially higher support among respondents with higher levels of concern across both treatment and control groups. Moreover, the effect of priming territorial loss is positive for evaluations of populist parties in opposition,  particularly for voters with low-to-moderate levels of concern. 
\begin{figure}[htp]
        \centering
        \includegraphics[width = \textwidth]{Figures/marginaleffectsall.png}
        \caption[{Marginal Effects of Loss Prime by Loss Concern}]{Marginal Effects of Loss Prime by Loss Concern\footnotemark}
        \label{fig:popratingint}
    \end{figure}    
Next, we turn to the question of how preexisting attitudes toward territorial loss shape the electoral payoffs of efforts to raise the salience of territorial loss narratives. To address this question, in Figure \ref{fig:popratingint}, we report the marginal effects of the\footnotetext{See Table C3 in the Online Appendix for full model specifications.} prime by territorial loss concern based on a quadratic interaction model.\footnote{See the Appendix for linear interactions and other specifications. In order to interpret this interaction effect as evidence of causal moderation, we must assume that the observed level of concern is unaffected by variation in question order. Since the measurement of territorial loss concern is itself the prime, this item is inherently post-treatment. In practice, there is no clear reason to expect reported concern to vary systematically due to question order, and there is no statistically significant average effect of the question's location on reported concern. Moreover, as we show in the next section, the reported concern is stable over time. Nevertheless, this caveat should be borne in mind when interpreting these results.} The figure confirms that, in line with $\mathbf{H}_4$, the effect of the territorial loss prime on nationalist populist party support was noticeably more positive for respondents with higher territorial loss concerns and only achieved statistical significance among respondents with moderate-to-high levels of reported concern.

    \begin{figure}[htp]
    \centering
    \makebox[\linewidth][c]{\begin{subfigure}[t]{0.45\textwidth}
        \centering
        \includegraphics[width=\textwidth,height=50mm]{Figures/marginaleffectsinpower.png}
        \caption{Parties in Power} \label{fig:popratingintpow}
    \end{subfigure}
    \begin{subfigure}[t]{0.45\textwidth}
        \centering
        \includegraphics[width=\textwidth,height=50mm]{Figures/marginaleffectsnotinpower.png} 
        \caption{Parties in Opposition} \label{fig:popratingintopp}
    \end{subfigure}}
    \caption[Marginal Effects of Loss Prime on Party Rating by Party Status]{Marginal Effects of Loss Prime on Party Rating by Party Status\footnotemark}
    \label{fig:popratingsint}
    \end{figure}
    
    However, Figure \ref{fig:popratingsint} also indicates an important difference in the way priming territorial loss affects respondents with high and low concern between nationalist populists in power and those in opposition. Whereas the prime has a positive but statistically insignificant effect on the respondents' evaluations of opposition parties at both low and high levels of concern, priming the respondents with territorial loss \textit{decreases} the rating of nationalist populist parties in power among those with low levels of concern, and has a significant positive effect only on those with moderately high levels of concern. 
    This pattern suggests the potential existence of a trade-off facing nationalist populists in power that does not affect their opposition counterparts. While priming territorial losses may increase support among those who view them as important, it has the opposite effect among those who do \textit{not}. While this finding falls outside the scope of our stated hypotheses, one potential explanation is that nationalist populists in power have strong ownership of the territorial issue, such that for those voters for whom it has a negative valence, priming the issue generates harsher evaluative judgments of the party.  It is also possible that mentions of lost territories are interpreted as a more credible threat of aggressive (and potentially costly) policy when made by parties that hold power, as opposed to marginal opposition parties. For those who view such outcomes as undesirable, then, being reminded of the association between governing nationalist populists and irredentist claims may carry more weight. While our data does not allow us to distinguish between these two explanations definitively, we contend that the latter is more likely, since there is no evidence in any of the cases under consideration that populists in opposition have weaker ownership of the territorial issue; if anything, the opposite is true with regards to both Jobbik in Hungary and the MHP in Turkey prior to its alliance with the ruling AKP \citep{fisher2009echoes,kovacs2013post}.\footnotetext{See Tables C4 and C5 in the Online Appendix for full model specification.} 



       
    %Consistent with H3, moreover, the governing AKP, despite attempts to capitalize on the issue of lost territories and an ongoing military operation in Syria at the time of the survey, actually received lower support among respondents who were more highly concerned about territory (though the effect fell short of achieving statistical significance) . Instead, higher concern over loss was associated with stronger support for the major opposition parties, the CHP and IYI, both of which have a history of emphasizing territorial disputes with Greece. Contrary to H3, however, Fidesz enjoyed dramatically greater support among respondents who were more concerned about lost territory. Although Orban has actively \say{played the territory card}, making particular efforts to mobilize the Hungarian diaspora, specific progress on the issue of lost Hungarian territory remains unlikely, suggesting that incumbents may be able to benefit from politicizing territorial loss despite an inability to take any action on it. It is possible that this result partly reflects the extremely strong partisan signals that Fidesz supporters receive, including through state-run media. However, it should be noted that the territorial loss prime did not have any significant effect on Fidesz ratings.
\subsubsection{Mechanisms: Retrospective Assessments of the Past}
    Although these results strongly support an association between concern for lost territory and support for nationalist populists, they provide limited information on the channels through which salient territorial losses impact voters' political attitudes beyond the narrow issue of irredentism.  While our theory suggests that salient losses matter politically because they emphasize a nostalgic view of an imagined better past, the evidence of a causal effect of salient loss on support for nationalist populists presented thus far does not allow us to evaluate this interpretation. In particular, we cannot rule out the potential alternative explanation that priming territorial loss increases affinity for nationalist populist parties because of issue bundling, without otherwise affecting political attitudes. 
    
    To explore this question, we included an item that asked respondents to name the best period in their country's history, chosen from a list of pre-specified periods delimited by regime change and/or major political upheavals such as wars.\footnote{For the full list of historical periods and distribution of responses see Table A.8 in the Appendix.} For all four countries, the historical period corresponding to the country's greatest territorial extent is also associated with a lower average standard of living and a less democratic regime (to varying degrees) than at the time the survey was conducted, as well as several other available periods. As such, while we cannot determine whether those who named these periods did so \textit{because} of a preference for greater territorial extent, the average propensity to do so can nevertheless be interpreted as a preference for a nostalgic past over the present. As such, if priming territorial loss induces respondents to identify the period of greatest territorial extent as the best time in their country's history at a higher rate, we interpret it as evidence consistent with our theoretical expectation that salient losses affect retrospective assessments of the past.
    
    
            \begin{figure}[!hbtp]
        \centering
        \includegraphics[width = \textwidth]{Figures/Historical Period Predicted.png}
        \caption[Predicted probability of naming the period of greatest territorial extent as the best time in the country's history]{Predicted probability of naming the period of greatest territorial extent as the best time in the country's history\footnotemark}
        \label{fig:terrexteeffs}
    \end{figure}


    Figure \ref{fig:terrexteeffs} shows the predicted probabilities of naming the period of greatest territorial extent across all countries by experimental treatment status and concern, while the results for individual countries and periods are reported in the Appendix. Consistent with expectations, both the prime and the level of concern are positively associated with a significantly higher likelihood of naming the period of the greatest territorial extent. However, the priming effect becomes insignificant for respondents with very high levels of concern, suggesting that there may again be a ceiling effect on the potential priming effect due to the already-high salience of the territorial issue among this group.
    
    Notably, these results also hold in Germany (see Appendix for country and period-specific results), where the territorial loss prime is associated with a large increase in the odds of naming the Nazi or Kaiserreich eras\footnotetext{See Tables C6 and C7 for full model specifications.} as the best period in the country's history. Given the ongoing association between nationalist populism and fascist legacies \citep{caramani2019national}, this strongly suggests the ongoing power of territorial aspirations to shape perceptions of the past, even in the least likely cases such as Germany, where irredentism has been broadly discredited and marginalized in mainstream political discourse.

 %   In order to test whether this tendency for respondents primed with territorial loss to adopt a more nostalgic view of their country's history accounts for its impacts on party evaluations, we also conduct mediation analysis, implemented using the R package \texttt{mediation} \citep{tingley2014mediation}. Under the standard sequential ignorability assumption (that is, we assume that the propensity of respondents to name the period of greatest historical extent as the best period in their country's history is as-if random after conditioning on treatment assignment and observable covariates, including loss concern), this allows us to separate the effect of the treatment into direct and causally mediated effects. Table \ref{tab:med} reports the results treating ratings of populist parties as the outcome, disaggregated again by parties in power and in opposition.

    


% Although there is no evidence of a mediated effect when considering all parties, the expected pattern again emerges when disaggregating by opposition status. Whereas populists in power suffer from a negative effect of nostalgia, populist parties in opposition benefit both from a direct boost from priming historical losses and a slight additional increase from nostalgia. These findings further support an interpretation of this divergence as deriving from dissatisfaction with government performance: whereas supporters of opposition parties may hold out some hope of a return to better times, this comparison is unfavorable for populist leaders in power who have been unable to bring about such a return in reality.

    
%     Finally, the effects of priming territorial loss on policy preferences towards the diaspora are reported in Figure \ref{fig:dias}. Across all cases, priming respondents with territorial losses slightly increased support for funding cultural programs, but \textit{decreased} support for exerting diplomatic pressure or imposing sanctions on average. This latter result is driven primarily by respondents in Romania, where the territorial prime was associated with a large decrease in support for these policies. This is likely due to the specific nature of Romania's loss, as hostile action toward Moldova (the most commonly named loss) is inconsistent with the goal of reunification. Only in Turkey did the prime increase support for the most extreme policy of increasing military funding, while it was slightly and nonsignificantly decreased in the other two countries. Despite this, many respondents \textit{do} support the more extreme policies in both countries: the average level of support even for military action is slightly over three on a five-point scale for both. More likely is that the treatment is simply not strong enough to elicit a substantial response on these more sensitive policy dimensions due to ceiling effects, as those who are willing to support them likely already do so, and those who do not are unmoved. \par 
% \begin{figure}
%    \includegraphics[width =\textwidth]{Figures/diaspora_comb.png}
%     \caption{Effect of priming territorial loss on support for policies towards the diaspora. Error bars show 95\% confidence intervals.}
%     \label{fig:dias}
% \end{figure}

\subsubsection{Extensions}

While the main analysis in this section has highlighted the overall impact of territorial losses on support for nationalist populist parties, and distinguished between ruling and opposition nationalist populists, in this subsection we briefly discuss two extensions, for which full results are available in the Online Appendix, that provide a more nuanced picture of the relationship.

The first extension returns to the theoretical discussion about the relationship between nationalism and populism in political party appeals and addresses the question whether the link between territorial loss and nationalist populist party appeals may be driven disproportionately by one of the two dimensions. To do so (see Appendix A), we replicated the main analyses using support for parties that meet only one of the criteria for nationalism or populism, but not both. While we find that higher territorial loss concerns also increase support for nationalist-only and populist-only parties, these effects were substantively smaller than for nationalist-populist parties, and fell short of statistical significance for populist-only parties. Furthermore, neither the experimental prime nor the interaction between the prime and loss concern yielded statistically significant effects for either nationalist-only or populist-only parties. While these results need to be interpreted cautiously given the small number of parties in these two categories, they nevertheless suggest that territorial losses are particularly conducive to strengthening political support for parties that combine nationalist and populist appeals.

The second extension explores the impact of the different types of territorial losses identified by our survey respondents and discussed in Section \ref{sec:maps}. Following \citet{cederman2024future}, we code each lost territory as corresponding to either a case of lost unity -- instances, such as Romania/Moldova or Hungary/Transylvania, where the dominant ethnic group was formerly a part of a single contiguous state but subsequently split into two sovereign entities, or lost home rule (i.e. other cases where control over territory is at stake but there is no ethnic splitting). The results presented in Appendix B highlight a few differences in patterns: in Romania (and to a lesser extent in Germany) the substantive effects of loss concerns on nationalist populist party support were larger for respondents who identified lost home rule territories, though the effects were still positive and highly significant for those who identified cases of lost unity. \footnote{In both of these cases, the instances of lost home rule are significantly less salient in public discourse than those of lost unity, such that one possible explanation for this pattern is that respondents with high levels of concern for these territories are simply among the most committed irredentists and wish to signal their attitudes by naming more obscure instances of territorial loss.} Meanwhile, in Turkey neither type of territorial loss had a positive effect for the nationalist populist incumbents but for the nationalist-only opposition party (iYi) we do find positive and significant effects for lost unity but not for lost home rule territories, a finding that is in line with the party's political platform. Finally, in Hungary, where all the territorial losses are coded as lost unity, we also find strong positive effects of loss concern on nationalist populist party support. Overall, the results in Appendix B suggest that concerns for both lost home rule and lost unity territories can provide electoral fuel for nationalist populist parties but that the relative strength of these appeals differs as a function of the particular historical and political context of different countries.   

    \subsection{Behavioral Consequences of Concern for Lost Territory: Voting for the AUR party}
    \label{sec:aurresults}

As discussed earlier, the sudden rise in popularity between consecutive waves of the survey of the AUR, a nationalist populist party openly running on a platform emphasizing reunification with Moldova as a high priority and criticizing the established parties over their inaction on that front, presents an ideal opportunity to examine the actual behavioral consequences of attitudes toward territory. Since the design of Study 1 involves asking \textit{all} respondents, regardless of treatment assignment, how concerned they are about lost territory -- as well as which specific territories -- at some point in the survey, it is possible to test for all respondents who participated in both waves whether those who were initially more concerned about the loss of territory were also more likely to defect to the AUR.

A potential concern in this analysis is the high attrition rate between waves, as the correlation between attrition and political attitudes could lead to spurious findings. Although Wave 1 territorial loss attitudes are uncorrelated with attrition, the potential for indirect confounding still exists. To mitigate this issue, we re-weight all respondents by the inverse of their predicted probability of attrition, calculated using a CART model on the full set of covariates.

\begin{table}[!htbp] \centering 
  \caption{Effects of Concern over Loss on AUR Support} 
  \label{tab:votaur} 
  \begin{adjustbox}{max width=\textwidth}
\begin{tabular}{@{\extracolsep{5pt}}lccc} 
\\[-1.8ex]\hline \\[-1.8ex] 
\\[-1.8ex] & Voted for AUR (W3) & Close to AUR (W3) & Loss Concern (W4) \\ 
\\[-1.8ex] & (1) & (2) & (3)\\ 
\hline \\[-1.8ex] 
Loss Concern (W1) & 9.995$^{***}$ & 22.921$^{***}$ & 0.590$^{***}$ \\ 
  & (3.745) & (4.078) & (0.065) \\ 
  COVID Restriction Support Index (W1) & $-$4.040 & $-$25.692$^{***}$ &  \\ 
  & (2.935) & (6.490) &  \\ 
  University Educated (W1) & $-$3.849$^{***}$ & $-$2.655 & 0.059  \\ 
  & (1.419) & (2.891) &(0.048)  \\ 
  Voted for AUR &  &  & $-$0.052 \\ 
  &  &  & (0.168) \\ 
 N & 278 & 277 & 164 \\ 
 Fixed effects & Yes & Yes & Yes \\
  Controls & Yes & Yes & Yes \\
\hline \\[-1.8ex] 
\multicolumn{4}{l}{$^{*}$p $<$ .1; $^{**}$p $<$ .05; $^{***}$p $<$ .01} \\ 
\multicolumn{4}{l}{All models include a series of attitudinal controls from the first wave, regional fixed effects as well as inverse probability attrition weights. See Table C9 in the appendix for complete results.} \\ 
\multicolumn{4}{l}{Model (1) is a logistic regression.}\\
\multicolumn{4}{l}{Loss Concern and COVID Approval are rescaled to range from $0$ to $1$.}\\ 
\end{tabular}  \\ 
\end{adjustbox}
\end{table} 

As the first two columns of Table \ref{tab:votaur} clearly show, the expectation that concern about lost territory causes defection to nationalist populist parties is strongly borne out in this case. For both actual voting and subjective proximity to the party measured after the election, the effects of lagged territorial loss concerns were statistically significant and substantively large. In model 1,  the effect of going from the minimum to maximum of territorial loss concern on a 100-point scale was roughly twice as large as the effect of having a university education, which is the other strongest predictor of AUR voting. 

Respondents' approval of the government's handling of the pandemic in 2020\footnote{Specifically, we use a five-item scale measuring respondents' degree of satisfaction with public health measures taken by the authorities, Cronbach's $\alpha=0.83$} (again, prior to the AUR's rise to prominence), moreover, \textit{does not} predict the decision to vote for the AUR. However, according to model 2, it is significantly associated with reported closeness to a substantively similar degree to territorial loss concern. Given the AUR's position as an anti-lockdown party amid the COVID-19 pandemic -- an aspect of its program frequently emphasized by political observers at the time \citep{rfe2020aur} -- this finding is especially remarkable. Although the small sample size implies that caution should be taken in generalizing this conclusion, it strongly suggests that, at least for a significant number of voters, latent concern over Moldova played as much of a role in defecting to this new nationalist populist party as did disillusionment over lockdown policies.

The main threat to interpreting these estimates as causal is the possibility voters are simply taking their cues from elites and that an (unmeasured) preexisting exposure to AUR leaders or other affiliated elites is generating the association, rather than a genuine effect of stable attitudes. While the timing of the waves makes this unlikely, were this the case, we would also expect a post-election effect of AUR partisanship, as people bring their views more in line with the party after voting for them. Column 3 of Table \ref{tab:votaur}, which regresses the reported concern in the fourth wave on voting in the third wave and concern in the first, reveals the opposite: attitudes are highly stable across waves, and having voted for the AUR has no statistically significant effect, with the point estimate in the opposite direction from expected. It is, therefore, highly unlikely that elite signaling drives the observed relationship, lending further weight to a causal interpretation.

\section{Discussion}
In this paper, we argue for the role of territorial loss as an important and, thus far, overlooked determinant of support for nationalist populist parties. Due to its close conceptual affinity with the rhetorical strategies employed by nationalist populists, the issue of territorial loss is particularly well-suited to parties of this type. By playing on a combination of nationalism, victimization narratives and nostalgia, nationalist populist parties in countries that have experienced significant territorial losses can benefit from positive associations with an imagined past while discrediting their mainstream opponents. While earlier work had highlighted the role of communist legacies in driving support for the radical right in Eastern Europe \citep{bustikova2009radical}, our paper suggests that we also need to pay attention to a different -- and often older -- type of historical legacy: territorial loss.

Using a combination of a panel survey and an original experiment, we provide evidence to support this intuition, demonstrating that attitudes toward territorial loss are robustly associated with party evaluations and vote choice. This holds observationally: individuals who report greater concern over past territorial losses also indicate greater support for nationalist populist parties. Our experimental findings, however, provide mixed evidence for a straightforward causal interpretation of this relationship, as respondents who reflect on past losses do not subsequently increase their support for nationalist populist parties on average. While this indicates that populist parties may have little to gain from simply raising the salience of past losses, it is important to note that the prime used in this article is quite minimal, consisting only of a factual question without any persuasive content. As such, we cannot evaluate the possibility that populists can mobilize support based on latent concern over territorial losses by framing them as part of a broader narrative of opposition to domestic and foreign elites, which would account for the robust observed association.

However, we find some evidence of heterogeneous effects of the experimental prime. Those with a moderately high level of concern over the loss they named, respond to the prime by evaluating nationalist populists more positively, indicating that nationalist populists seeking to mobilize support around feelings of loss may benefit primarily through its heightened salience for those who already feel strong attachments. At the same time, analysis of heterogeneity between governing and opposition parties indicates that populists in power face a trade-off between gaining support from this demographic and losing it from those for whom territorial losses carry limited weight. 

At the same time, descriptive evidence suggests that the magnitude of this trade-off may be relatively small. With the exception of Germany, the distribution of concern over territorial losses skews rather high in all cases despite the limited scope for territorial aggression in Romania and Hungary. Notably, the distribution of concern reported in Hungary and Romania is not significantly different from that in Turkey  -- a country currently occupying parts of neighboring Syria. To the extent that populist politicians stand to gain from exploiting popular concern over past territorial losses, our results suggest that they frequently enjoy a deep well of potential support.

This interpretation is supported by our analysis of a natural experiment created by the emergence of a new populist party in Romania between survey waves. The case of the AUR allows us to demonstrate that attitudes to territorial loss are both stable over time and have a causal effect on voting for populism, and not the reverse. While politicians undoubtedly play a role in fostering such attitudes over the long term, the results presented in this paper are more consistent with the interpretation of historical attitudes as stable features of individuals that are shaped by education and social environment than as something that can be created from thin air by opportunistic elites. Instead, at least over the relatively short time scale considered here, our findings suggest that the association between territorial loss concern and support for populism stems from the ability of populist parties -- especially in opposition -- to exploit existing attitudes, rather than to create them.

Beyond the narrow issue of support for populist parties, we find large effects of the minimal prime of lost territory on evaluations of national history, including a significant rise in positive evaluations of the Nazi era in Germany. Unlike party ratings, which may be relatively settled, people are rarely called upon in everyday life to determine the greatest period in their country's history. In this context, the effect of reflecting on loss is to increase the relevance of territory in historical evaluations. This, in turn, has potentially dramatic implications for contemporary politics: if political entrepreneurs successfully equate the value of historical models with their territorial extent, it may be easier to convince their audience to accept trade-offs regarding redistribution, regime type or other relevant policy dimensions.

Taken together, the results of this paper suggest that, despite strong international norms to the contrary, territorial claims remain an important determinant of domestic politics, and should not be overlooked. Even in cases such as Germany, where the idea of territorial expansion has been repudiated in mainstream political discourse, large numbers of people continue to feel at least some concern over past losses. This latent concern represents a potential wellspring of support for populist challengers, even if they face potential trade-offs concerning alienating voters who are unconcerned about territorial losses. As such, future analyses of the sources of support for nationalist populist parties would benefit from considering the potential for mobilization around historical narratives of territorial loss.

Given that the results in this paper are based on evidence from a set of four -- albeit diverse -- European countries, it is important to address the scope conditions of our argument. On the one hand, the four countries considered here have suffered rather substantial territorial losses during the 20th century, which raises the possibility that the mobilizational potential of territorial loss appeals may be weaker elsewhere. On the other hand, a large and diverse set of countries worldwide have experienced territorial losses, and the politicization of these losses has both triggered interstate conflicts and  reverberated in domestic politics. 

Russia's invasion of Ukraine, the dispute over the Malvinas/Falkland Islands between Argentina and the United Kingdom and China's claims on Taiwan are among the most visible and explosive recent instances of territorial loss politics, but there are many other historical and contemporary instances. It is worth also considering the timing of territorial losses: while, for the four countries we analyzed in this paper, the losses date back roughly 75-100 years, there are a number of countries that suffered more recent perceived territorial losses, including India, Indonesia, Somalia, Sudan, Ethiopia, Georgia, Moldova, Serbia, and, more recently still, Armenia and Ukraine. While our findings suggest that territorial losses have long political half-lives, we would nevertheless expect the intensity of their political reverberations to be greater in countries with more recent territorial losses. Finally, the dynamics of territorial loss-based political appeals will depend on the nature of political competition in a given country. The evidence in this paper suggests fairly similar electoral dynamics across a set of regimes that range from liberal democracy (Germany) to competitive authoritarian regimes (Turkey and, arguably, Hungary), but it is less clear how they would apply in regimes where incumbents exert even stricter control over electoral competition, such as Russia or China. 

Perhaps the most important scope condition of our theory, however, relates to the presence of nationalist populism. While territorial nationalism \textit{per se} is a near-universal phenomenon in contemporary politics -- albeit stronger in some places than others -- populism has generally been thought to be characteristic primarily of Europe and Latin America. In Latin America, however, populism often takes on a more inclusionary character, focusing on social justice and anti-elite rhetoric rather than on nationalist issues \citep{mudde2013exclusionary}, suggesting that the pull of nationalist populism may be weaker. Even in these contexts, there are instances where territorial issues intersect with populist narratives. For example, the territorial disputes and the question of national sovereignty in Argentina, Bolivia and Chile over maritime access have  played significant roles in shaping political discourse. Given the recent rapid increase in nationalist populism in Europe and the United States, moreover, it is quite plausible that similar dynamics might arise in other parts of the world where latent potential for the politicization of territorial loss exists. We hope that future research can address these scope conditions more thoroughly.  










\end{flushleft}

\newpage
\printbibliography
\end{document}
    